DSC Europe 23
2023-11-20..2023-11-24
Belgrade, Serbia
| title | author | length |
|---|---|---|
| Whose job does AI automate? | Cassie Kozyrkov | 52m |
Claim: modern ML reduces complex automation to two instructions — optimize this goal on that data set — which shifts work from writing line‑by‑line code to… | ||
| From Data Lakes to Data Mesh - The Evolution of Redmesh | P. Ilic, S. Riling | 30m |
Bosch Digital describes Redmesh as a hybrid self‑service platform that turns centralized data lakes into domain‑oriented data products, using orchestration… | ||
| Using Semarchy to Master Data Management | Cvetanka Eftimoska | 21m |
Semarchy is presented as a flexible MDM platform that centralizes customer/product records, performs exact match and fuzzy match consolidation (examples use a… | ||
| Improving YouTube Recommender systems for big screen users | Rein Zhang | 24m |
YouTube runs a multi-stage recommender that serves billions and is shifting from 2015 linear models through multitask DNNs to large contextual models in 2023… | ||
| Bringing AI to Your Data | Tomasz Czarnecki, Vanja Melnik | 31m |
Google Cloud demonstrates concrete paths to production: TPU v5 hardware claims 4x performance, Vertex AI combines Vertex AI Search and Vertex AI Conversation… | ||
| Intro to GKE setup of airflow with helm and terraform | Ved Prakash | 31m |
Provisions GKE with Terraform, deploys Airflow 2.5.3 using Helm charts, and runs Airflow on private VPC with Cloud SQL Postgres and a git-sync sidecar for… | ||
| Retrieval Augmented Generation | Spela Poklukar, Tea Brasanac | 31m |
Retrieval Augmented Generation combines embeddings and vector search to cut hallucinations by feeding an LLM retrieved, up‑to‑date snippets; the demo connects… | ||
| Our way to a data-driven enterprise - A Journey by Bosch | Rainer Metje, Wolfgang Klein | 45m |
Bosch describes a concrete data transformation: a federated Red Mesh with an API enablement program and an internal marketplace launched in May that lifts API… | ||
| The Art and Science of Infographic Design | Maxim Gorbachevskiy | 29m |
Practical steps to turn analyst output into action by climbing the DKY pyramid (Data→Information→Knowledge→Wisdom) and reducing decision risk, shown with a… | ||
| Panel: Main legal risks and challenges with LLMs / generative AI models | Marija Vlajkovic, Rania Wazir, Heiko Hotz | 55m |
Panelists map concrete legal pain points: models train on scraped social media and personal data so 175 billion parameters scale leaks consent and opt-out… | ||
| A Journey of Auditlogs from Kafka to ElasticSearch | Muhammad Arslan | 26m |
Built a centralized audit pipeline that writes Avro-compressed events to MSK Kafka, uses a Confluent connector to bulk index into OpenSearch and concurrently… | ||
| Visual Design - Do not let Data be your cul-de-sac! | Rutuja Pawar | 27m |
Claims that 9 out of 10 companies get stuck in a “data dead end” because dashboards show insight but don’t lead to decisions, and prescribes a workflow: brief… | ||
| Official Opening – Alex Linc | Alex Linc | 07m |
Conference claims rapid growth from a 250‑person local start nine years ago to events in Belgrade, Vienna, Zagreb, Miami and planned Cairo and Paris, running… | ||
| Official Opening – Nicola Pontanara | Nicola Pontanara | 06m |
World Bank outlines concrete Serbia programs: a portfolio of 14 projects / $1 billion funds public-sector modernization and private innovation, a $20 million… | ||
| Official Opening – Edward Ferguson | Edward Ferguson | 09m |
UK digital sector employs over 3 million people, contributes £150 billion a year, has a combined market cap above £1 trillion and counts 144 unicorns; the UK… | ||
| AI-Driven Business Success: Measurable Results and Revolutionized Sales | M. Grbovic | 29m |
Bank moves sales from opinion to data driven systems, using eightfold more data over five years aiming 10x, with 1.2 million customers and 120,000 SMEs as the… | ||
| Optimizing Text Classification in Construction Data | V. Jovanovic, K. Arandjelovic | 28m |
Over a million tender paragraphs with ~22,000 potential products, ~100 categories and suppliers, and ~300 positions per tender require a month of manual… | ||
| Transforming the Search Industry with LLM Using Source Frontier | Miodrag Vladic | 20m |
Claims a production information retrieval layer called Source Frontier that already stores tens of billions of vectorized text segments, integrates with an LLM… | ||
| Augmented Retrieval Products with GAI models | Marcel Tkacik | 35m |
Describes building retrieval‑augmented products that run on Azure OpenAI and Azure Cognitive Search, deploying a prototype called project Guru and a commercial… | ||
| Data storytelling in the era of generative AI | Diego Marano | 27m |
Claims a working service that uses Large Language Models over clustered consumer surveys to create evidence-informed synthetic personas that answer marketers… | ||
| Responsible AI: Rewriting or Repeating History? | Alexandra Ebert | 36m |
History shows concrete failures and fixes: photographic tech takes 110 years until color photos serve darker skin, car safety goes from red flags to airbags… | ||
| Chromaphone Analysis in Multimedia: Big Data Formula | P. V. Kocovic, M. Ramachandran | 32m |
In their talk, Petar & Muthu delved into the concept of a "light formula," a mathematical representation elucidating the transformation of the raw energy of… | ||
| ML in Production on Databricks | Petar Zecevic | 29m |
Focuses on concrete production steps: use Delta Lake for ACID files, schema enforcement and time travel; run transformations with Delta Live Tables to get… | ||
| Using AI for cheaper and more efficient solar power | Milan Bradonjic | 27m |
End-to-end pipeline that turns a single aerial image into panel layouts and yearly yield estimates using rot net for roof alignment and multiple segmentation… | ||
| Bosch DataOps: NLP Development on Kubernetes | Igor Stranjanac, Filip Novovic | 29m |
Bosch extends DevOps into DataOps/MLOps by running a Kubernetes-based platform on AWS that supports ~30 researchers, a 5M-entity / 10M-relationship knowledge… | ||
| Mastering AI Issue Debugging: A Data Scientist's Guide | Liliya Akhtyamova | 28m |
Demonstrates an automated debugging pipeline that ties New Relic metrics, S3-stored JSON payloads and a SQL document store to a single Trace ID lookup, letting… | ||
| From Psychohistory to Curious Machines | Luca Morena | 48m |
Company Nextatlas scrapes major social media, cleans noise, and scores profiles to surface early adopter identification rather than influencers, using an… | ||
| Cognitive Agents for Primary School Education: Visual Approach | Bogdan Okresa Djuric | 33m |
Primary students draw and name colorful avatars that are turned into interactive cognitive tutors that answer spoken curriculum questions in Croatian and reply… | ||
| Deep learning module recognizes video sounds | Pawel Ekk-Ciernikowski | 22m |
Project pipelines automated transcripts and background-sound labels using Azure Video Indexer for timestamps, Azure Databricks for custom ML and orchestration… | ||
| Facebook data and machine learning predicting elections? | Alen Kisic | 22m |
Hand-collected Facebook data over three months from 41 French cities covering 225 candidates feeds 25 variables (normalized per voters) into four ML methods… | ||
| An AI assistant for football tactics | Petar Velickovic | 34m |
An AI system called TacticAI models corner kicks with graph neural networks and geometric augmentation, predicts first-touch receivers with 78.2% top three… | ||
| Bridging Gaps: Classic Approaches Beyond Deep Learning Constraints | Sead Delalic | 25m |
Practical cases show when classic methods beat heavy models: document classification uses contour detection plus pixel/histogram comparison instead of… | ||
| Adversarial Attacks | Aleksandar Tomcic | 21m |
Covers concrete attack types and limits: gradient-based minimal perturbations (DeepFool-style, L1/L2 norms) that flip a panda to a gibbon with imperceptible… | ||
| Revealing the Unseen: Visualizing Attention in Transformer Models | Catalin Hanga | 24m |
Explains how tokenization and high‑dimensional embeddings feed into query, key, value projections and how dot‑product self‑attention produces context‑aware… | ||
| Panel: Open Data In Public Use | M. Gajic, A. Glusevic, S. Verbic | 47m |
Panelists report Serbia publishing GTFS feeds for six cities and restoring Google Transit after a deletion, build daily-refresh APIs for transit, ran… | ||
| Make data central feature | Antoni Ivanov | 24m |
Argues that modern tools shift from managing tasks to modeling data sets and shows concrete demos: DBT for SQL models, a Versatile Data Kit notebook that… | ||
| Opening up the box: the complexity of human interaction with AI systems | Rania Wazir | 23m |
Argues that every AI design choice matters: image prompts in stable diffusion default to whiteness unless explicitly corrected; a chatbot (ChatGPT) fabricates… | ||
| Data science in high energy physics | Mila Pandurovic | 33m |
Overview of a CEPC Higgs analysis that reconstructs a six-jet H→WW hadronic final state using WHIZARD and PYTHIA for event generation and CEPC ILD detector… | ||
| Networks Cocaine and Fruit Flies | Milan Petrovic | 23m |
Group experiments use plexiglass arenas and a CTech flight‑tracker to record 30 adult male Drosophila (3–5 days old) across 20 bottles, comparing regular food… | ||
| Panel: Gender bias in AI - In collaboration with Women in Data Science Belgrade | Siyu Sun, Isidora Gataric, Ljubica Vujovic Boskovic | 55m |
Panelists describe concrete interventions and failures: a racing game removed gender field so analysis of male vs female behavior disappears; a lending model… | ||
| Custom chips needed for efficient high-performance computing | Mihailo Isakov | 24m |
Argues that single‑thread gains are exhausted and custom silicon is the practical path for latency‑sensitive, energy‑bound workloads, backing it with analysis… | ||
| Applied Custom Vision | Catalin Gheorghiu | 29m |
Shows how to turn existing RTSP camera streams into actionable signals by training Azure Custom Vision models (roughly 25+ labeled images per class), deploying… | ||
| Generative AI: Impact of Responsible AI | Shahab Anbarjafari | 26m |
Covers why generative AI shifts risk from research to production and what concrete controls look like: traces the model leap to Transformers (Dec 6 2017)… | ||
| Guiding AI Governance Evolution: Traditional to Generative AI | Therese Svensson | 21m |
Claim: 86% of Swedish CEOs plan to deploy generative AI within 12 months and foundation models are often built by tech giants scraping web-scale data then… | ||
| Under the Hood: Key Algorithms Powering Vector Search | Dusan Milunovic | 32m |
Covers HNSW, LSH and product quantization with concrete tradeoffs: HNSW + PQ yields fastest searches with ~0.86 recall after ~50s build time while flat scan is… | ||
| Decoding AI Landscape: Business, Market Trends, and Risks | Goran S. Milovanovic | 01h 03m |
A data‑driven audit of generative AI that claims we already have a system‑one superintelligence (excellent at associative tasks) but not system‑two reasoning… | ||
| How our neural networks find polygons to understand your documents better | Boris Zimka | 31m |
Pipeline targets 4‑point ID crops and 6‑point passport contours and evaluates methods by critical error rate and IoU. A lightweight segmentation network… | ||
| Engineering Intelligence through Agent Interaction | Igor Sevo | 29m |
Transforms RPA into agentic process automation by combining GPT‑4 Vision and vector/semantic indexing into chained, policy‑based agents that keep source… | ||
| How to Build and Scale Ultra-Lean Gen AI Startups | Tarry Singh | 53m |
Asserts that multiplying human creativity by GenAI (an M² mental model) beats chasing giant training clusters; shows a lean stack instead: a toolkit that does… | ||
| Robotic emotions | Max Talanov | 31m |
A bio-plausible mouse-brain model reproduces dopamine, serotonin and noradrenaline subsystems and maps Hugo Lövheim’s cube of emotions to cortical activity… | ||
| Beyond Deep Learning: The Cognitive AI Frontier | Rana Gujral | 46m |
Claims a shift from pure deep learning to cognitive AI that fuses machine learning, NLP, pattern recognition and emotional-signal processing, with concrete… | ||
| Finetuning Stable Diffusion with low-rank adaptation techniques | A. Kovalchuk | 32m |
Claims that low-rank adapters reproduce near full fine‑tune quality while cutting memory and trainable parameters: Stable Diffusion’s UNet is ~860M params… | ||
| UAV-based Person Detection and Geolocation for Search and Rescue | M. Ivasic-Kos | 30m |
Pipeline builds a custom UAV dataset of about 2,000 images captured with a Phantom 5, heavily augmented to simulate snow rain and occlusion, then retrains YOLO… | ||
| Seeing Through the Lens of Transformers: A New Era in Computer Vision | I. Biliskov | 33m |
Vision Transformers replace convolutions by splitting images into patch-based tokens, flattening and linearly projecting them, adding position embeddings, then… | ||
| Redefining User Experience with Large Language Models | Igor Ilic | 46m |
Copilot is an Office‑embedded assistant grounded in your documents that pulls context from Microsoft Graph and uses OpenAI models to generate drafts, slides… | ||
| Your data lakes need architecture too! | Tomislav Hlupic | 28m |
Covers concrete architectures and tradeoffs: the original two-layer architecture (landing + raw) for cheap cold storage, Bill Inmon’s Data Pond split… | ||
| Real-time analytics at IoT scale | Pramod Immaneni | 35m |
Rivian runs a cloud streaming stack on AWS EKS that ingests telemetry into Kafka, Flink, Druid and persists segments to S3, claiming sustained 2 million… | ||
| Robust CI Workflow with dbt for Data Warehouse Optimization | Matteo Molteni | 19m |
Midsize company (100 employees, ~50 engineers) with a two-person data team builds a dbt-centered CI that enforces git-only changes, dynamic per-model Airflow… | ||
| Scalable Data Flow Management for Smart Farming Digital Twins | Mihailo Ilic | 27m |
Claims a working interoperability stack that feeds actionable digital twins to reduce chemical use (aligned with EU targets of 50% pesticide and 20% fertilizer… | ||
| Simplified AI-Powered Market and Consumer Research | M. Bjeloglav, V. Vujacic | 29m |
A production Brand Pulse integration turns plain‑English prompts into real analyses by mapping language to documented variables, calling APIs and executing… | ||
| Panel: LLM use in business transformation | J. Djurovic, M. Milivojevic, S. Dejanovic | 44m |
Panelists report a deployed fraud tool that analyzes 30 million transactions/day and say conversational-transcript summarization and smart search lift… | ||
| Enhancement of Demand Forecasting and Supply Chain Planning product via GPT | N. Pozar | 22m |
BE-terna extends its Demand Forecasting and Supply Chain Planning product by embedding a GPT-based chatbot and a supply chain monitor that ingests supplier… | ||
| Why New Products Fail: Harnessing Sales and Marketing in Post-Human Era | P. Reinisch | 24m |
Case study of a 50‑person, €5M AI vendor shows what changes: product roadmap restructured from playground to POC to production, POCs budgeted as deliberate… | ||
| Proptech Innovations: Data-Driven Strategies for European Properties | Ivan Livic | 27m |
VendueTech aggregates 250,000+ EU auctions and builds an analytics pipeline on Azure + Postgres + Elastic to surface distressed properties with algorithmic… | ||
| The Innovator’s Dilemma: to Build or Not to Build | Danijela Horak | 30m |
Midsize companies face a clear choice: adopt off‑the‑shelf LLMs or invest in shaping models with their own data. Evidence cited: a live BBC bird‑watching… | ||
| Impact of NLP in AI-powered Bank | Branislava S. Todorovic, Mirella Miskic | 33m |
NLB deploys two email classifiers in production since March: a 10-topic classifier and a sentiment pipeline built from 5,000 Slovenian emails with manual… | ||
| From Tables to Answers: building QA System for In-Document Searches | Vladimir Ageev | 28m |
Case study of a table QA pipeline that fine-tunes Tapas (base and large) on 30,000 annotations across ~3,000 tables to map queries to highlighted cells and… | ||
| Beyond GenAI | Jepson Taylor | 49m |
Argues that GPT-4 plus self-directed internal review can invent algorithms that outperform human-designed optimizers on benchmarks (red bars = human; non-red =… | ||
| Industry NLP: Multitask & Continual Learning | Tin Ferkovic | 28m |
Compares BERT‑based baselines (single‑task and shared encoder) with adapters and hypernetworks using three classification and four NER datasets on a single… | ||
| The iPhone Moment of AI | Lars Rinnan | 51m |
Argues generative AI hits an iPhone moment and shows concrete wins: decoding bee waggle dances to remap hive locations and stop local dieoffs, predicting… | ||
| AI in Media: Balancing Innovation, Quality, and Responsibility | Dominic Herzog | 29m |
TX Group applies models to auto-annotate archives and measure newsroom metrics so editors can set KPIs, for example tracking visibility of women rising 16% to… | ||
| New Extractivism | Vladan Joler | 01h 03m |
Reverse engineering maps one device (Amazon Alexa) to its data centers and patents, traces packets through national ISPs and mass pings of Serbian IPs, and… | ||
| From Algorithms to Assets: Data Science Meets Real Estate | Ilija Lazarevic | 28m |
Describes building automated valuation models (AVMs) that use historical transaction and listing data with a market‑approach (not forecasting) to price… | ||
| Capturing and Validating Real World Movements of Millions of People | Runxin Li | 34m |
Foursquare explains how owned apps power validation and produces 9 billion stops per month from 500 million devices, uses a simple provider-evaluation rule… | ||
| Ethical Considerations in Predictive Analytics | Bunmi Akinremi | 28m |
Covers concrete steps for ethical predictive analytics: prioritize robust real-world test sets, use explainability (feature attribution and attribution maps)… | ||
| AI and Geopolitical Strategy | Dragan Vitorovic | 34m |
Research backed by HFRI frames AI as a general‑purpose strategic enabler that behaves like finance and like nuclear deterrence, enabling covert weaponization… | ||
| Responsible AI 2.0 : NO AI without RAI | Sray Agarwal | 15m |
A concrete roadmap for Responsible AI that insists on end-to-end governance: a checklist of copyright and attribution, bias detection, explainability, privacy… | ||
| Brain-Computer Interfaces: Ethical issues and resolutions | Kevin LaGrandeur | 34m |
BCIs combine implanted or scalp sensors with AI-triggered stimulation to stop seizures or abort depressive episodes; NeuroPace-style implants sense onset and… | ||
| Panel: Can AI transform business? | T. Zwingmann, V. Sivasubramanian, A.Kalam | 42m |
Panelists argue that AI is an enabler not a magic wand, showing concrete pilots: developer copilot trials with ~10% early adoption, customer-care assistants… | ||
| "Unit tests for Data" with Deequ | Pavle Tabandzelic | 39m |
Demonstrates Deequ running on Apache Spark with a declarative API plus small user lambdas to enforce concrete constraints—completeness, uniqueness, ranges—and… | ||
| Deploying computer vision models at scale devops-free | Ricard Borras Navarra | 22m |
Veriff moves from one-model-per-microservice to serving many models from a shared GPU pool using Triton Inference Server packaged into model repos and stored… | ||
| Approach to Architecting Generative AI Solutions | Ivan Petrovic | 32m |
A practical blueprint for building generative AI systems that names concrete components: use vector databases for semantic search, separate user and agent… | ||
| Under Pressure: Applying ML in Real Time | Miroslav Bicanic | 24m |
A real-time ML pipeline is built to hit 100 transactions/sec with <=100 ms latency on a single moderately powerful machine by iterating tools and architecture… | ||
| How is it to lead an all-remote, globally distributed team? | Dennis Van Rooijen | 27m |
An all‑remote data platform runs on Snowflake and Airflow, executes about 250k queries/day, processes ~2,700 daily models and runs 88 active Airflow flows… | ||
| Evolving Technology Excellence; Dual Leadership Experience.. | G. Savic, S. Mrsic | 35m |
A digital service company describes a concrete org and delivery playbook: a multi-level hub-and-spoke model plus a three-stage governance maturity chart… | ||
| Leading with Data: Crafting an Effective Strategy for 2023 and Beyond | Luka Anicin | 32m |
Argues that an executable data strategy starts with a prioritized roadmap, a data asset inventory and quick wins before building custom models, using templates… | ||
| Growth Hacking - the next best thing to transform digitally? | Nikolay Nikolov | 24m |
Argues growth hacking must be a company-wide program with measurable foundations rather than a siloed marketing trick, using RFM model demo logic, transition… | ||
| The AI-Based Banking Era | Gianluca Berghella | 26m |
Company-built AI platforms run in over 200 banks and a 200-person team focuses on integrating machine learning across front office, back office and compliance… | ||
| How to screw Agile and digital transformation | Davor Gasparac | 34m |
Argues that Agile gets wrecked by one‑size‑fits‑all rollouts, consultants selling Scrum = Agile, and teams organized around IT systems instead of business… | ||
| How Data Hub flies above the clouds! | Strahinja Miljkovic, Nikolaos Servos | 45m |
A centralized Data Hub built in 2020 ingests social media and enterprise sources into Azure Data Lake using the Delta Lake format, processes with Databricks… | ||
| Data Lakehouse: Who, Why and How | Milos Solujic | 34m |
Explains why merging lakes and warehouses fixes silos and lowers cost, naming S3 + Parquet for storage and Delta Lake and Apache Hudi/Iceberg as the… | ||
| Cloud warehouse monitoring ( Snowflake case study ) | Josip Saban | 37m |
Snowflake case study shows a monitoring system built from account logs and Airflow metrics that identifies runaway compute: a single Excel-style warehouse… | ||
| AI Ops and its challenges | Alexander Buyanov | 24m |
Claims a unified AI Ops layer that merges CMDB data, observability telemetry and configuration management to control hardware (power, cooling), schedule… | ||
| How to Fail with AI - and How to Succeed? | Tobias Zwingmann | 41m |
Explains why projects fail: hyped expectations, leaving AI to IT, hiring a lone data scientist who leaves, and messy data quality that stalls results. Gives a… | ||
| Zero-party data giving AI the power of "why" | Umar Akhtar | 19m |
Kudo collects zero-party data via targeted surveys, builds AI-generated personas and a conversational Chat to Persona interface, then pushes segments into a… | ||
| Creating a Data Platform for Software-Defined Electric Vehicles | J. Shiverick, V. Sivasubramanian | 24m |
Rivian’s data platform ingests telemetry from 65–70k vehicles, decodes packet captures in Go, and publishes Parquet/Delta on S3 to a Lakehouse holding 12… | ||
| Eye-tracking and AI: Unveiling the Limitations and Synergies | Milica Kovac | 20m |
Company-built webcam-based remote eye tracking and a portable lab tracker (Essa/Tobii) are compared against two popular predictive saliency models across… | ||
| Process mining from the inside | Oksana Astapova | 34m |
Process mining turns event logs into an auto‑discovered process map and interactive KPIs, letting teams simulate changes and spot bottleneck shifts… | ||
| Unlocking Geospatial Data Science: A Strategic Overview | Milan Janosov | 30m |
Concise survey of geospatial data science: Python-based workflows (matplotlib, GeoPandas, Pydeck, Folium) plus big-data stacks on Databricks Mosaic and Sedona… | ||
| Therapy's Role in Easing Anxiety and Decision-Making | Mihailo Joksimovic | 28m |
Personal account centers on 7+ years of REBT therapy that follows a burnout and shows measurable changes: an average 11 month delay between suggestion and… | ||
| How to treat your data as a product | Ivan Dundovic | 29m |
Argues for shifting ownership left so domain teams deliver and own reusable datasets, citing the four data mesh principles and recommending data mesh thinking… | ||
| How can we use AI to bring new value to the business | Nenad Bozic | 27m |
SmartCat claims more than 20 solutions in production and shares a concrete roadmap: discovery workshop, ideation paired with technical due diligence, quick… | ||
| Cloud Advisory: Cloudy, with a chance of people | Josema Cavanillas | 48m |
Argues that cloud migrations fail less for tech than for people, using the Hittite collapse of 1190 BCE as a concrete analogy and insisting on discovery first… | ||
| AI Engineer Ethics: Shaping Responsible Generative AI | Heiko Hotz | 43m |
Shows concrete defenses for generative AI: deploy Llama and Falcon endpoints behind request filters (keyword rules, business rules or a classifier like… | ||
| Building successful data teams in the realities of 2023 | Karl Ivo Sokolov | 29m |
Data teams shrink while expectations rise: budgets tighten, Snowflake bills shock finance, and 85% of data projects fail (Gartner), so prioritize work that… | ||
| Bridging Business and Tech for Data Science Success | Nina Mrzelj | 22m |
Claims that 87% of data science projects fail and pins most failures on weak cross-team communication rather than algorithms. Recommends concrete fixes: run… | ||
| Asynchronous Work: The Next Phase of Remote Work | Radovan Bacovic | 37m |
GitLab runs 1,600 people in 56 countries serving 30+ million users with zero offices and a 2,500‑page public handbook that documents their async operating… | ||
| Why Governments Should Build Their Own AI Products | Vid Stimac | 33m |
Government-run project builds an AI platform with 32 A100 GPUs and an MLOps team, funds an Institute for Artificial Intelligence, and embeds a delivery unit in… | ||
| Lean Experimentation and Decision Science with Humans in Warehouses | Sayon K. Saha | 32m |
A picker-scheduling system combines stacked regression time-series forecasts with a mixed-integer programming linear optimizer to recommend shifts that match… | ||
| Change the world of logistics through addHelix | Claus Vogelmann | 20m |
addHelix claims concrete savings by cutting kilometers and truck count through route optimization using genetic algorithms and ETA prediction with neural… | ||
| Retail Transactions at scale for AI - From PoC to production | Ayoub Fakir | 24m |
A production pipeline ingests roughly 1 billion transaction files per year from 2,000+ stores across 70+ countries, converts ugly XML with 340+ fields into… | ||
| Advanced search for your e-commerce website | N. Bozic, N. Jankovic | 31m |
Shows that replacing keyword search with semantic search via vector embeddings (GPT and Hugging Face models) yields immediate relevance gains, handles… | ||
| The specifics of the AB-test for Unit Economics with a network effect | A. Chernobrovov | 26m |
Randomized A/B on per-order unit economics breaks when shared fixed costs make group outcomes interdependent: a variant can look better per-order in the slice… | ||
| AI-Powered Transformation in Fashion E-Commerce | Ratko Nikolic | 24m |
Claims a measurable lift: cites McKinsey figures that 35% of Amazon purchases and 75% of Netflix views come from recommendations and marketer stats of ~56%… | ||
| The Power of Personalized Customer Experiences Through AI Innovations | N. Trajkov | 28m |
Describes a practical build of a recommender for a sports retailer: data aggregated with pandas, features encoded with count vectorizer, similarity scored via… | ||
| Panel: AI impact on E-commerce: Friend or a Foe? | B. Klasnja, M. Bukorovic | 47m |
Panel details operational wins and open problems: a mathematical route model combines sales visits and deliveries and already trims routes (about 10 trucks per… | ||
| Data Science in Gaming | Siyu Sun | 49m |
Gameloft describes concrete systems for monetization and measurement: an in‑game IP shop with targeted offers, dynamic pricing tradeoffs, and… | ||
| Leveraging Sentiment and Topic Detection Analysis for Decoding Player.. | N.Vasiljevic | 30m |
Uses a mixed pipeline: start with manual qualitative coding on a 1,000–2,000 comment sample, then fine‑tune a token‑based NLTK sentiment intensity analyzer and… | ||
| ..Stories About Industrializing In-Game Machine Learning Solutions | Pavle Petrovic | 32m |
Ubisoft builds a modular Machine Learning Framework that moves inference into the engine to stop Python-over-sockets lag and let prototypes ship as game code… | ||
| AI – Legal Insights | Andrea Radonjanin, Nikola Kliska | 28m |
Explains legal realities not theory: US relies on fair use but recent lawsuits from Getty and class actions challenge scraping, EU uses the data mining… | ||
| WIPO's work on AI, Video Games, & Intellectual Property | K. Gygax | 14m |
WIPO runs a four‑year old Frontier Tech division and a twice‑yearly WIPO Conversation on IP and Frontier Technologies that draws over 5,000 people from 162… | ||
| Generative AI & Intellectual Property Law | Gaetano Dimita | 32m |
Generative AI in games creates two legal questions: who is the human author and is the output original under local law. Practical points: models like… | ||
| Unlocking revenue: newcomer classification for boosted profits | Dmitrii Matveev | 24m |
Built a multi‑task neural model with hard parameter sharing that takes ~300 features from a 1‑day observation window to jointly predict churn probability (BCE)… | ||
| Panel: AI Intellectual Property | Djordje Krivokapic, Gaetano Dimita, Andrea Radonjanin, Nikola Kliska, Ksenia Gygax | 52m |
Serbia reports creative industries at 5.5% of GDP with about 100,000 workers and video games downloaded/paid over 100 million times in 2022; the IP office uses… | ||
| Blockchain solutions: Ready for business | Tanja Bivic Plankar | 23m |
Blockchain adoption focuses on concrete wins: 420 million crypto owners and merchants reporting a 32% ROI lift and 40% new customers after enabling crypto… | ||
| How blockchain can enable a better and more ethical AI | Ida Pandur | 32m |
Claims that blockchain gives immutable on-chain identity and audit trails to trace deepfakes and model updates, with concrete builds: a facial emotion face… | ||
| Crypto tokenization of financial markets and real world assets | Aleksej Misailovic | 20m |
Claims that tokenization will reshape finance hinge on concrete pilots and numbers: major banks and exchanges run pilots (examples include JP Morgan digital… | ||
| Securing Merchants: Combatting Scams and Fraud for Trust | Ana Maria Anic | 33m |
Company Moonay runs an internal ML model plus multiple third‑party fraud vendors and manual reviews, uses wallet screening vendors to flag known scam… | ||
| Token Engineering Toolkit: Practical Insight | Aleksandar Damjanovic | 27m |
Claims practical workflow for tokenomics: ideation workshops, benchmarking, mathematical spec, then iteration with Monte Carlo, parameter sweeps and agent… | ||
| Intro to crypto derivatives | Nikola Krivacevic | 30m |
Explains that crypto derivatives now dwarf spot (Q2 2023 derivatives volume ≈ six times spot) and that perpetual futures account for roughly 90%+ of crypto… | ||
| Rewarding every sustainable commute with a new purpose-driven token | Marina Miletic | 20m |
A mobile app plus on-chain token that uses an ML movement validator to detect commute mode, calculate avoided CO2 versus a combustion car, and trigger a smart… | ||
| Tax issues of cryptocurrencies & NFT | Boris Subotic | 20m |
Starts from 2008 and traces practical tax pain points through the 2013 and 2015 bull runs, showing how authorities and taxpayers clash over definitions and… | ||
| Unleashing the Power of Data-Driven Crypto Investing | Stevan Radonjanin | 29m |
Built a crypto investment app that offers four risk tiers and two choices per tier, limits the universe to the top 50 coins, and uses CoinGecko data plus a… | ||
| Panel: The Hidden Value behind Cryptoanalytics | John Lilic, Johannes Pfferer, Srdjan Vukmirovic, Ana Maria Anic | 37m |
Cryptoanalytics uses the public ledger to detect scams and trace hacks faster than legacy finance — examples cited include Sam Bankman‑Fried’s rapid unmasking… | ||
| The Dawn of the Social Fintech; AI, Sustainability and Social Science | Luis Seco | 46m |
A practical roadmap linking AI, social science and finance: large language models power a product called responsibly that scores companies on… | ||
| AI - Digital (r)evolution in banking | Miroslav Marinkovic | 29m |
Banks already deploy AI as risk early warnings, biometric KYC, and chat GPT for CRM but no regional bank yet gives a truly personalized automated offer; boards… | ||
| Data Science in Asset Management | Ana Cukic | 16m |
Claims that the $100 trillion industry still runs on 30-year-old tech and redirected post‑GFC IT budgets to compliance, that many managers surrendered to… | ||
| Recent approaches in retail banking marketing optimization | Michal Ambler | 28m |
Banks generally run dozens to hundreds of propensity models (logistic regression and simple deep nets) on fragmented customer 360s, then feed scores into an… | ||
| Revolutionizing Financial Data Integration: The OpenBB Edge | I. Radovanovic | 28m |
OpenBB claims a modular V4 platform that moves beyond a single terminal into an extensions-driven ecosystem with a web API and a Python SDK, shedding beta on… | ||
| Digital central bank money: Ideas, initiatives and issues | Martin Summer | 49m |
Many central banks are moving from debate to design: the ECB publishes a legislative blueprint that treats a CBDC as digital euro as legal tender, distributed… | ||
| Digital, data-driven financial advisory | Mathias Frey | 36m |
Prototype asks for income and costs, hits Python models that take 7s (optimized to 1.5 seconds but still unacceptable), and collapses under load for 10 million… | ||
| Analyzing Daily Bank Transactions: Maximizing Insights | Jovan Milovanovic | 25m |
Claims that daily bank transactions spot problems months before annual accounts, backed by a product called heartbit that parses raw statements, maps flows to… | ||
| AI and Blockchain Shaping The Future | Ana Nives Radovic | 22m |
Claims a working approach that pairs smart contracts with AI to automate and adapt supply‑chain decisions, using blockchain immutability to protect training… | ||
| Model Validation Genesis | Urban Susnik | 30m |
Model validation in a legacy bank runs hundreds of models live and expands beyond checklists to changing regulations and the applications themselves… | ||
| AI in the bancassurance paradigm | Carlo M. Speranza | 37m |
Company builds a process-driven bancassurance platform called Arunda 360 that configures life and non‑life products across multiple carriers via configurable… | ||
| Insurtech – from papers to metaverse | Ivan Bimbi | 18m |
LAQO launches as a 100% digital insurer in September 2020 and sells a car policy with 20 clicks to buy, expands to travel and pet cover, and pilots… | ||
| Panel: Data Analytics In Investment Banks | Adam Milenkovic, Marko Jevremovic, Vlada Aleksic, Filip Bovan | 37m |
Panelists describe concrete wins and limits: inconsistent inputs produce valuations like $1B vs $1.5B for the same company, cloud migrations and unified… | ||
| How may AI help you? - healthcare | Ligia Kornowska | 35m |
Poland and the CEE region show concrete AI adoption: a curated list grows to 150 startups with 60% using AI but only about 6.6% of hospitals running AI in… | ||
| Hospital Network Advisory: Listen to your metadata! | Josema Cavanillas | 29m |
Describes inserting a passive ear on hospital LANs that sends only metadata to a cloud AI which reconstructs topology, counts devices, fingerprints OS and… | ||
| Addressing Social Isolation in Elderly Inbetweeners | Milos Todorovic | 19m |
Claims a concrete platform: AIO two-sided marketplace matches verified companions with “elderly Inbetweeners” for home visits, companions submit an 18-point… | ||
| AI and Blockchain Framework for Healthcare Applications | Muthu Ramachandran | 34m |
Introduces a concrete Secure and Sustainable Software Engineering Framework for Healthcare Blockchain Applications, S3EF-HBCA framework, with practiced… | ||
| Clinical Data Management and Clinical Data Science A Practical Approach | U. Vilmanovich | 28m |
Company builds an integrated EHR/EDC platform (beta at Switzerland’s largest hospital) that captures 5–10,000-question forms, links wearables and SmartWatches… | ||
| AI-Enhanced Medical Mask Compliance in Healthcare | J. Bubonja, A. Vucicevic | 25m |
Project releases a benchmark of 10,000 images with over 50,000 annotated people and ~35,000 masks present, labeled for mask fit, mask type and scene metadata… | ||
| Big data in personalized medicine | Branka Zukic | 17m |
A Western Balkan genomic survey uses whole genome sequencing on regional cohorts to identify drug–pharmacogenomics marker pairs and to design a preemptive gene… | ||
| Data-Driven Approach in Treating a Medical Condition | Goran Dumic | 20m |
Parents log every diaper, medication, meal and symptom into synced cloud storage, build dashboards from an ETL pipeline and a 350-column table, try… | ||
| AIMED: AI for Prostate Cancer Detection and Immunotherapy Response | T. Krizan | 30m |
Two EU-funded programs build a non‑invasive prostate screen around a urine vaporization device that produces spectral profiles with up to 6,000 omic features… | ||
| Machine Learning Techniques for Omics Data Analysis | Vesna Pajic | 29m |
Team uses ctrp portal gene-expression and drug-response data to predict cell-line sensitivity, starting from 686 samples and after cleaning using 516 (train… | ||
| Data-Driven Insights for Drug Kinetics & Dynamics in Therapeutic Strategies | K.Vucicevic | 31m |
Shows concrete results: fits nonlinear mixed-effects population modeling to clinical and real‑world PK/PD data, quantifies interindividual, interoccasion and… | ||
| Enhanced EEG Data Classification with 3D SNN and On-Line Learning ESN | Dimitar Penkov | 13m |
Hierarchical model maps 14-channel EEG (60 one-second trials, 128 time bins) through a NeuCube 3D spiking layer with STDP synapses whose per-neuron firing… | ||
| HealthLy app Digital solution in Healthcare - Data Capture | Haris Elhag | 28m |
HealthLy is a family-focused digital therapeutics app that claims to raise adherence (noting ~50% of patients miss doses) and collects 85+ parameters across… | ||
| Computer Vision in Melanoma Detection | Djordje Hirs | 25m |
Uses public ISIC/HAM10k-style sets with strong class imbalance and common artifacts (rulers, hair, gel bubbles) that can leak labels; notes one dataset had… | ||
| Network neuroscience: where the brain and social networks works the same | Ilya Zakharov | 26m |
Claim: brains and social networks share Rich Club hubs that concentrate communication and produce short paths (Facebook measured ~3.57 handshakes). Methods… | ||
| A Sustainable Future for Learning AI in Healthcare | Dimitrios Kalogeropoulos | 50m |
A sustainable path rests on concrete fixes: a systematic review finds only 10 RCTs out of 8,000 supervised deep‑learning studies, radiology accounts for 70% of… | ||
| Do Androids Dream of Human Doctors? | Ognjen Milicevic | 31m |
Shows real failures and edge cases: noisy ECGs and Holter outputs that miss or flag issues (one trace yields 90 candidates), a pregnant ECG where fetal… | ||
| SMART HEALTH HOME: Technology, healthcare, legal, and socioeconomics | Vladimir Brusic | 48m |
A working Smart Health Home prototype connects wearable sensors through an edge computer + mobile hub to a data exchange center, records nearly 10,000 hours of… | ||
| Panel: AI-Enhanced Healthcare: Future Trends | Mohamed Ramzi, Ognjen Milicevic, Nevena Veljkovic, Nebojsa Sabovic | 52m |
Panel lays out concrete friction points and wins: Universities lack basic math and Excel literacy while industry needs NLP/LLM engineers; regulators and… | ||
| Applying AI for Threat Detection in the World of Cybersecurity | Aleksa Stojanovic | 22m |
Claims that AI systems at MDS informatic engineering convert signature based gaps into zero-day detection without signatures, process logs for real-time log… | ||
| AI Impact on Cybersecurity and Privacy | D. Pleskonjic, V. Jelic | 48m |
Claims that GitHub Copilot output contains 40–45% vulnerable code and that AI-trained network models reach >99% accuracy after a few weeks guide the talk… | ||
| Implementing AI in Critical systems | Branko Primetica | 44m |
Organizations face a shortage of high‑quality labeled data and real interoperability, so success requires mandating data standards and building large clean… | ||
| Unlocking security with Seif.ai | Uros Arsenijevic | 22m |
Seif.ai presents a three-part platform: employee training center with customizable evolution courses, nano‑learning and phishing simulations tied to individual… | ||
| Building Digital Trust | Sanja Kekic | 01h 17m |
Claim: digital trust is the next stage after digital transformation and needs an operational digital trust framework (ISACA-influenced) that covers onboarding… | ||
| An open dialogue about the state of crypto, especially Ethereum | John Lilic | 40m |
Argues that Ethereum has effectively become a global USD banking system thanks to stablecoins and wallets, not the originally promised “world computer,” and… | ||
| Demolishing AI Gatekeepers | James Corbett | 18m |
Claim: Big Tech blocks AI progress by locking training data, GPUs and talent behind walled gardens and quotas — startups often get ~10 GPUs and VCs back… | ||
| Clustering players' strategies for generation of non-player characters | Rok Piltaver | 29m |
Clustering turns millions of recorded formations into hundreds-to-thousands of playable templates by splitting data by formation size, applying a two-step… | ||
| Improving Cybersecurity Posture with an AI Security Assistant | Goran Gvozden | 26m |
An AI security assistant that automates mapping of CVE-style vulnerability descriptions to MITRE ATT&CK’s 14 tactics and 600+ techniques, enriching each… | ||
| AI in HCI: To Serve and Protect | Alesandar Jevremovic | 31m |
Describes a funded project called Casper that builds a working prototype semantic firewall to block harmful content for vulnerable users by combining… | ||
| Panel: Humanity in the Loop: Balancing AI, Data Science, and Human Ingenuity in the Cybersecurity.. | Vlatko Canev, Marko Elazar, Vladan Todorovic | 50m |
Panelists describe concrete use of ChatGPT-assisted reporting and quick scripting (one example: a custom password-generator that produces 200–300 GB in one… | ||
| Fireside chat | Branko Primertica | 44m |
Focus on practical failures and fixes: critical infrastructure AI breaks because of poor or insufficient data, incompatible vendor formats, and attacks that… | ||
| Gamification within Oil and Gas Industry | Nikola Stojanovic | 13m |
A VR training game demonstrates seismic interpretation by importing OPEX seismic cubes via Python, converting the 3D matrix into a Minecraft-like 3D voxel… | ||
| Augmenting our Reality with Technology: A Practical Guide to Real-world.. | N. Sagir | 01h 10m |
Control Art builds real-time AR experiences on iPad and Epson smart glasses using Unity, ARKit/ARCore and custom shaders; projects run for 5–8 years with a… | ||
| Procedural this, procedural that | Nikola Damjanov | 34m |
Builds node-based pipelines in Substance Designer and procedural rigs in Houdini + VEX to automate textures, VFX and level dressing, shipping a… | ||
| Use of AI-powered tools for immersive experiences | Milan Licina | 33m |
Milan delved into the transformative realm of AI-powered tools for immersive experiences in his talk. Focusing on the intersection of artificial intelligence… | ||
| Why is data analysis more important in VR games than in traditional games? | A. Djura | 34m |
VR games generate dense spatial and body-tracking signals so analytics must catch posture, gaze, hand joints and exit behavior rather than just clicks. Key… | ||
| Maximizing Data in Gaming IP with GRID Platform | Dominika Szot | 31m |
GRID ingests official game-server events and exposes live, historical and aggregated feeds via a title-agnostic SDK that maps common datapoints so new FPS or… | ||
| Integrating Stable Diffusion Generative AI in Artist Workflows | I. Jovanovic | 29m |
Uses the open-source pipeline centered on Automatic1111 + ControlNet with upscalers, hypernets and textual inversions/LoRA trained from ~10–20 images; runs… | ||
| Centaur Drawings | Uros Krcadinac | 30m |
An exhibition recreates algorithmic co‑creation by pairing a pen plotter drawing robot with human visitors and a homebuilt LSTM in Python that generates asemic… | ||
| Re-imagining mobile gaming studios with Gen AI | Abhijeet Katte | 18m |
jult applies generative AI to ship features faster and cut costs: a level-design translator that localizes and generates puzzles across languages and saves… | ||
| Introduction to data mining (especially for healthcare domain) | Milena Stojic | 01h 23m |
Hands-on demo using the Orange GUI to build a binary heart‑disease classifier on a cleaned dataset reduced to 14 attributes (originally >76). Data come from… | ||
| Data Radar Maps: Navigating Product Growth with Data Insight | Joy Chatterjee | 25m |
Data Radar Maps are a spider-chart framework that scores an org across organization, taxonomy, tools and processes, fluency and utilization to reveal exactly… | ||
| ChatBot Learning Assistant with Large Language Models | Vladislav Belov | 32m |
Built a production MVP in five months: a learning chatbot that answers course questions with retrieval-augmented generation, returns chapter references and… | ||
| Revolutionizing Automation: Retrieval-Augmentation with LLMs | Djordje Grozdic | 32m |
Shows a working RFP automation app that parses uploaded PDFs/Word/Excel, splits text with overlapping chunks, embeds chunks into a vector store… | ||
| A Hybrid LLM based Enterprise Chatbot Solution Weaver Peer by Intellya | A. Petrovic | 31m |
Hybrid system merges an intent-and-scenario chatbot with retrieval-augmented generation and LLMs to handle questions reliably while covering edge cases… | ||
| Innovating the Telecom Industry: Embracing Automation and AI | Nebojsa Dobrijevic | 18m |
Oron Telecom (third largest Serbian operator, claims largest download speeds locally) presents concrete projects: a speech-to-text → LLM → TTS pipeline for… | ||
| Introduction to orchestration ETL processes with Apache Airflow | Miroslav Tomic | 01h 59m |
Hands‑on demo shows a reproducible cluster: a Dockerized Airflow cluster with MS SQL source and Postgres target, GitHub code and 400MB backups. Concrete… | ||
| Relation extraction from texts | Andrija Poleksic | 01h 30m |
Walks through building a relation-extraction pipeline that uses OpenNRE and fine-tunes BERT-base-uncased on a curated slice of a unified distant‑supervision DB… | ||
| Some Implications for the use of Gen AI in courts | Thiago F. Avanci | 53m |
Generative LLMs like ChatGPT/GPT-3.5 are already appearing in real court work: judges in Brazil and Colombia use them to draft rulings, one case contains false… | ||
| From RNNs to Transformers: A Journey through Language Modeling | J. Bubonja | 01h 26m |
Covers tokenization choices (character, word, BPE/subword), shows word embeddings trained via co‑occurrence that produce word2vec analogies, explains recurrent… | ||
| LLM applications with LangChain | Andrej Miscic, Luka Vranjes | 01h 59m |
Hands-on demo builds three concrete apps: a QA bot over internal docs using FAISS vector store and OpenAI Ada embeddings (26 source pages → 346 chunks), a… | ||
| Transaction log as the heart of Delta Lake | Boris Matijasevic | 01h 21m |
Delta Lake stores every change as numbered JSON commits (names use up to 20 zero‑padded digits) in a transaction log alongside Parquet files so Spark replays… | ||
| Complex networks sparsification: methods and tools | Marija M. Dankulov | 01h 06m |
Walks through sparsifying a NIPS paper–word bipartite network and shows concrete effects: starting graph is ~1,500 papers and ~16,000 nodes with ~1.1 million… | ||
| Molecule Testing for Ansible-Driven Data Lake Deployments on Kubernetes | J. Triva | 01h 21m |
Shows how Molecule runs automated Ansible tests against Vagrant/VirtualBox instances to build a six-node Kubernetes cluster on a single VM, provisioning master… | ||
| Build your own Generative AI Chatbot with Watsonx Assistant | Erik Ternav | 01h 54m |
Step-by-step demo builds a working chatbot using Watsonx Assistant + Watsonx.ai LLM, connects webchat via one-line embed, and wires RAG through Watson… | ||
| Optimizing workloads with Transformers and Diffusers | Stefan Dejanovic | 01h 16m |
Shows concrete speedups and exact steps: using Hugging Face Transformers pipeline yields summarization in 42s, switching to Optimum cuts it to 27s; converting… | ||
| Analyze your data in Python | Isidora Gataric | 01h 53m |
Step‑by‑step Python workflow: Jupyter Notebook with pandas + Jupyter + Plotly to load fake Excel/CSV data, inspect dtypes, drop or impute NaNs, and detect… | ||
| Object detection using OpenCV | Amir Siddiqui | 01h 11m |
A hands‑on OpenCV tutorial that loads pre‑trained Haar Cascades in a Jupyter notebook to detect face, eyes and smile with visible bounding boxes, shows… | ||
| Google Colaboratory and Python: Deep Learning basics for CV tasks | Kristina Host | 01h 55m |
Demonstrates how to build, train and test a simple CNN end to end using Google Colaboratory, Python, OpenCV, matplotlib, TensorFlow and Keras, including saving… | ||
| Introducing Two Criteria for Future Data Prediction with ML Models | D. Radivojevic | 01h 09m |
Defines two concrete pass/fail rules for next-step prediction: distance-domain criterion requires model metrics to beat average differences of random pairs… | ||
| Fall in Love with Data Transformation: DBT as Your Weapon | R. Bacovic | 01h 42m |
Hands‑on demo installs DBT with Poetry, connects to Snowflake (or DuckDB locally), loads sample tables and runs transformations that create prep and prod… | ||
| MRI Brain Scan Classification Guide | FastSurfer, XGBoost, 3D CNNs | D. Polak | 01h 20m |
Step-by-step pipeline shows extracting volumetric features with FastSurfer + XGBoost pipeline (FastSurfer segments a brain in ~1 minute versus ~10 hours for… | ||
| Build Your Own AI Chatbot with OpenAI & LangChain | Christophe Zoghbi | 01h 12m |
Workshop shows a runnable pipeline that indexes local docs, embeds chunks and answers queries using OpenAI API and LangChain. Text is split into 17 chunks… | ||
| Hands on Quantum Computer | Petar Korponaic | 01h 36m |
Builds a working Variational Quantum Eigensolver that finds the minimal eigenvalue of a 4x4 Hermitian matrix (two qubits) and matches the classical result… | ||
| Create ML framework - introduction to no-code ML | Aleksandar Petrovic | 01h 29m |
Builds working models in minutes using Create ML with drag‑and‑drop export to a .mlmodel and direct import into Xcode; trains a tabular classifier for shopping… | ||
| Building & Optimization of Neural Networks from Scratch with Numpy | M. Marjanovic | 01h 25m |
Builds a working neural network from scratch using a NumPy-only implementation, coding sigmoid activation, its derivative, and the backpropagation equations… | ||
| Model validation using deep generation of stress data | Vitaliy Pozdnyakov | 01h 01m |
Uses a weekly US gasoline dataset to stress-test a forecasting pipeline: a LSTM baseline (input size 10 weeks, horizon 100 weeks, backtesting error ≈0.24–0.25)… | ||
| Shinylive - Running R and Python Dashboards without a Server | Mario Annau | 01h 44m |
Demonstrates running full Shiny R and Python dashboards entirely in the browser by compiling runtimes to WebAssembly and using webR/pyodide plus a… | ||
| Unleash the Crypto Data Landscape in your Hands | Maxim Legg | 01h 33m |
Hands-on workshop exposes an open index of Uniswap V3 price and liquidity with two million blocks (16M–18M) of Ethereum mainnet data (about 13M swaps, ~628k… | ||
| Fundamentals of Deep Learning Workshop | Abdelghani Kabot | 01h 30m |
Hands‑on workshop walks through building and training models from scratch with a first lab on MNIST handwriting digit classification, then explains training… | ||
| Transforming DataFrames: Intro to Graphs & GNNs with PyTorch Geometric | D. Cvetkovic | 59m |
Tutorial demonstrates converting tabular data into graphs with PyTorch Geometric and NetworkX and trains simple GNNs on concrete datasets. It loads a Spotify… | ||
| The Synthetic Data Advantage: How It Helps You to Get Both AI Privacy & Explainability | A. Ebert | 01h 00m |
Creates synthetic structured data on the Mostly AI platform and runs notebooks in DataCamp to test utility using a Paris Airbnb set (~56k rows, 80/20 split)… | ||
| Productionizing Jupyter Notebooks | Antoni Ivanov | 01h 18m |
Demos a workflow that turns exploratory Jupyter notebooks into runnable data jobs using Versatile Data Kit (VDK): notebooks are tagged as VDK sales to enforce… | ||
| Automation of the Automation using ChatGPT | Gennadii Chursov | 01h 16m |
Walks through generating Playwright + TypeScript API tests with ChatGPT (stable GPT‑3.5) via Postman/API, saving generated files to a GitHub GPT branch… | ||
| Gamified Reinforcement Learning: Mastering AI through Video Games | L. Jovanovic | 01h 47m |
Builds a simple Flappy Bird gym in Python using pygame + numpy and trains a population with a genetic algorithm (no backprop). Agents use a 4-input MLP with 32… | ||
| Simplifying Event-Driven Services with FastStream: A How-To Guide | Tvrtko Sternak | 51m |
FastStream is an open‑source Python framework that wires Pydantic models to async brokers so handlers are called on incoming messages and return values are… | ||
| Deep Learning Techniques for Monocular Depth Estimation | Vladimir Kovacevic | 01h 30m |
Covers three concrete systems with code and Colab notebooks: a transfer‑learning U‑Net using a DenseNet169 encoder with 42M parameters and a loss mixing pixel… | ||
| Creating Ethical Generative AI for Enterprises | J. Skalec, M. Pap, D. Doljalnin | 01h 21m |
Hands-on demo of IBM Watsonx.ai shows building enterprise chatbots that use Watson Assistant plus Watson Discovery for OCR/KB search and RAG with embeddings… | ||
| Augmented Reality - ARKit framework showcase | Marko Mutavdzic | 53m |
A small iOS app that applies face filters using Apple ARKit and SceneKit loads USDZ/USDC assets, positions and rescales them as child nodes of a face anchor… | ||
| Identification of Building Thermal Model - Using Data | Mihael Jaksic | 48m |
Claims a data‑driven pipeline replaces year‑long semi‑physical calibration and the Unscented Kalman Filter by training lightweight models on short windows of… | ||
| Build End-To-End MLOps Platform with Open Source DVC Ecosystem | Mikhail Rozhkov | 02h 10m |
Step‑by‑step demo of an end‑to‑end MLOps stack that actually runs: git + DVC for data and model versioning (dvc init / dvc add / dvc push / dvc pull), dvc.yaml… | ||
| Lasso and Ridge regression models in R: Introduction | Milica Maricic | 02h 10m |
Hands-on R demo using the WEF dataset, glmnet and cv.glmnet to show that regularization reduces overfitting: ordinary lm has adjusted R² = 0.182, Ridge shrinks… | ||
| Storytelling through data | Milica Scepanovic | 01h 10m |
Shows that poor scaling and excess styling make viewers answer basic questions incorrectly, citing an eye‑tracking KPMG study and a corrected example that… | ||
| Panel: Can Generative AI be infused in Gaming? | M. Jevremovic, N. Damjanov, M. Licina | 45m |
Panelists argue generative AI is already in games — used for textures and environment population, procedural asset generation and even whole prototypes — and… | ||
| Revolutionizing AI Adoption with No-Code Efficiency for AI & ML Development | M.Rozhkov | 25m |
Describes practical no code and low code steps: use DVC for data and model versioning, DVC X for generative AI scale to billions of files, drag‑and‑drop UIs… | ||
| AI's Impact on Croatia's Digital Healthcare Transformation | Anja Baresic | 41m |
Croatia has e-prescriptions since 2011 and e-referrals since 2020 and now runs a 16-entity consortium with 3 million EUR fund to build an AI-first digital… | ||
| Data Governance in Large Enterprises | Biljana Atanaskovic, Petra Hajdukovic, Miha Kovac, Andrej Hudoklin, Aleksandar Nedeljkovic | 01h 20m |
Retail and FMCG groups report concrete results: a three-year data strategy, a cloud single source of truth on Azure, and a virtual Center of Excellence that… | ||
| Research and Innovate the Future of Interactive Entertainment | Konrad Tollmar | 40m |
EA research describes concrete tools and results: Swish real-time cloth ML replaces hand‑made simulation in shipped games, improving quality; light Stitch… | ||
| Interview with Vladimir Brusic | Vladimir Brusic | 06m |
Describes rapid adoption of wearables and cheap Internet of Things sensors for continuous health measurement but stresses that the key problem is… | ||
| Interview with Vladan Joler | Vladan Joler | 09m |
Argues that AI cannot reliably solve social problems because training data always encodes human choices and no perfect dataset exists, and compression in… | ||
| Interview with Spela Poklukar | Spela Poklukar | 06m |
Daily use of GPT yields a roughly ~50% hallucination rate that drops for popular libraries but spikes for niche tools (example given: MLflow). Retrieval and… | ||
| Interview with Rania Wazir | Rania Wazir | 02m |
Argues that diplomacy must move responsible systems from guidelines into enforceable practice by pushing responsible and trustworthy AI across borders, not… | ||
| Interview with Rana Gujral | Rana Gujral | 04m |
Defines cognitive AI as a self-learning adaptive model that mimics human thought by reasoning over unstructured data rather than relying on fixed rules, and… | ||
| Interview with Philipe "Indy" Reinisch | Philipe "Indy" Reinisch | 03m |
Claims that nobody yet knows the most efficient way to employ AI in diplomacy and that the immediate task is learning how to use emerging technologies and AI… | ||
| Interview with Milan Janosov | Milan Janosov | 03m |
Claims that geospatial work fails because people treat GIS and ML as separate worlds are front and center, arguing the real task is connecting GIS to… | ||
| Interview with Milan Gospic | Milan Gospic | 02m |
Claims that AI will provide data analysis and decision support over vast global datasets, flagging errors and supplying actionable context for policymakers… | ||
| Interview with Mathias Frey | Mathias Frey | 06m |
Banks already run production AI for anti-money laundering and fraud prevention, but broad advisory automation lags because “financial well‑being” is hard to… | ||
| Interview with Luis Seco | Luis Seco | 06m |
AI exposes baked-in biases by acting as AI as a mirror, surfacing behaviors society might not notice and forcing policy and design choices; technical… | ||
| Interview with Lars Rinnan | Lars Rinnan | 08m |
Argues that cinematic doomsaying overstates real risks while researchers' calls for pauses after GPT-4 reflect worry about unexpected uses; notes a one‑year… | ||
| Interview with Konrad Tollmar | Konrad Tollmar | 03m |
Game development converges on concrete tools: motion capture (MoCap) is used to make in‑game characters move more naturally, EA integrates it into production… | ||
| Interview with Ilija Duni | Ilija Duni | 03m |
Describes using Foursquare’s places and movements datasets in a FEMA collaboration to detect safe zones after disasters by spotting where people gather or stop… | ||
| Interview with Alexandra Ebert | Alexandra Ebert | 03m |
Claim: no model is unbiased but the target is to identify and remove unwanted or unintentional biases, not chase impossible neutrality. Practical route uses… | ||
| Aftermovie | 02m | |
A short aftermovie made from opening footage of DSC Europe 23 collates the welcome line "I would like to welcome you all" with onstage music and applause into… | ||
| nothing matches | ||