DSC Europe 24
2024-11-18..2024-11-22
Belgrade, Serbia
| title | author | length |
|---|---|---|
| Aftermovie | 02m | |
The video’s content is essentially a single vocal sample — the word "Heat" — looped and layered repeatedly, so the piece functions as single-word loop and… | ||
| Solutions prototyping using Azure AI Studio | Pawel Ekk-Cierniakowski | 01h 10m |
Shows rapid prototyping with Azure AI Studio using Speech, Vision, Language and Azure OpenAI: builds a chatbot backed by Azure AI Search / embeddings and… | ||
| Unlocking Personalization: A Deep Dive into Modern Recommendation Algorithms | Sarang Gupta | 01h 21m |
Walks through implemented pipelines on MovieLens 1M (≈1M ratings, ~6k users, ~3.8k movies) with runnable GitHub notebooks, showing heuristics (popularity… | ||
| Introduction to orchestration ETL processes with Apache Airflow | Miroslav Tomic | 01h 18m |
Demonstrates a Dockerized Airflow stack (webserver, scheduler, workers) orchestrating ETL between MS SQL Server and PostgreSQL using Python hooks and… | ||
| From Walruses to Code: A Guide to Tackling Unique ML Problems | Dmitrii Solomonov | 01h 09m |
Argues that niche vision problems reach usable performance without huge corpora by combining transfer learning with synthetic data generation and lightweight… | ||
| Data Engineering in the AI Era: Building a Robust Foundation for AI Applications | Artemiy Kozyr | 01h 17m |
Walks through a production stack that uses Debezium CDC into Kafka, Snowflake as the warehouse and Snowplow for event tracking, with DBT models, Kafka… | ||
| Building Multiagent Systems: An Ontology-Driven Approach with SPADE | Bogdan Okresa Djuric | 01h 24m |
An ontology-to-code pipeline translates a domain ontology into runnable agent templates for SPADE, generating agent classes, roles and behaviors (cyclic… | ||
| Tools and data to decarbonize cities: Citiwatts, CoolLIFE & an EU-27 Mapping of Financing Schemes | Giulia Conforto, Mostafa Fallahnejad, Aadit Malla | 01h 28m |
Open‑source platforms deliver runnable planning: CityWats (Hotmaps evolution) maps heat and cooling layers, runs district‑heating potential and dispatch… | ||
| Generative AI for Businesses | Erik Ternav | 01h 03m |
Shows IBM watsonx.governance and the watsonx.ai studio running in demo: prompt lab (chat, structured templates, freeform), foundational model library with… | ||
| Covariance-based & Partial Least squares SEM modelling: Outline & applications | Milica Maricic | 01h 42m |
Comparison shows concrete choices: R code uses lavaan and semPLS in R to fit CB‑SEM and PLS‑SEM, reads an SPSS file, computes Cronbach’s alpha >0.7 for scales… | ||
| Building Smart Agents with LlamaIndex: A Comprehensive Overview | Amir Siddiqui | 01h 06m |
Shows a working Jupyter/React agent built with Llama 3.1 that actually calls tool plugins (calendar, weather API, finance, DuckDuckGo search) and falls back to… | ||
| Automation of the Automation using GenAI | Gennadi Chursov | 01h 26m |
Generative AI calls the ChatGPT API to produce API tests, page objects and locators, using Playwright + TypeScript and Axios, saving code and execution logs… | ||
| How to build AI Agents: A step by step guide to creating AI agents | John Robert | 01h 30m |
A three-agent travel planner is built and demoed: a destination analyst that web-scrapes flights and weather, a city guide that gathers attractions and prices… | ||
| Supercharge Your Real-Time Analytics: Snowpipe Streaming & Dynamic Tables in Snowflake | V. Prakash | 01h 20m |
Live demo builds a real-time pipeline that injects row-level events directly into Snowflake using Snowpipe Streaming and transforms/aggregates them with… | ||
| HPC application in analysis of large scale complex networks | Marija M. Dankulov | 01h 17m |
Shows how Networkit on a GPU (tested on Google Colab T4) loads a ~YouTube dataset of about 1.1M nodes and ~3M edges, computes degree distributions (power‑law… | ||
| Comparing GNN Approaches for Molecule Identification | Ivana Milutinovic | 01h 20m |
Tutorial compares GNN pipelines that turn ≈11,630 SMILES into graphs, augment the minority classes to ≈20,800 graphs by node‑index permutations preserving… | ||
| Graph Based Approach to Text Mining Solutions | Stefana Janicijevic | 01h 31m |
Method demonstrates discovering new senses for Slovenian collocations by building a knowledge graph from a CLARIN.SI collocation table (~1,700 rows yielding… | ||
| Mastering RAG Evaluation with MLflow & Open Source LLMs | Cyril Kondratenko, Mikhail Rozhkov | 01h 28m |
Shows a reproducible RAG evaluation pipeline that uses MLflow, LChain, and open‑source models hosted in AI Studio to generate and judge QA pairs, deploys an… | ||
| Using Financial Data APIs: A Guide to Collection & Integration | Tamer Khraisha | 01h 18m |
Step-by-step demo that clones a GitHub repo, runs a Dockerized Jupyter notebook and uses Python/pandas to call five real APIs (openFigi, permID, Alpha Vantage… | ||
| Getting started with Containers and Kubernetes: A Beginner's Workshop | Rigerta Demiri | 01h 41m |
Step‑by‑step demo showing Docker basics and a tiny end‑to‑end example: writes a Dockerfile to build image, builds and runs public nginx images, then… | ||
| From Pixels to Words: Exploring Multi-Modal AI with Python | Nandana Sreeraj | 01h 04m |
Live demo shows a Colab Python workflow that loads CLIP (ViT‑B/32) via transformers and torch, preprocesses an online image with PIL/BytesIO, and scores three… | ||
| Building Data Breach Search Engines with MariaDB | Lukas Vileikis | 01h 23m |
Runs breachdirectory.com as a real data breach search engine built on MariaDB and InnoDB, handling billions of rows by looping queries across per-table files… | ||
| Google Colaboratory and Python: Deep Learning basics for CV tasks | Kristina Host | 01h 39m |
Hands-on Colab walkthrough using Google Colaboratory with GPU access, OpenCV and matplotlib for image I/O and augmentation, and TensorFlow/Keras to build a… | ||
| Data-Driven Game-dev Business: Building a Profitable Portfolio | Nikola Vasiljevic | 25m |
Builds a genre-ranking pipeline that aggregates sales, Metascore and user score, marketing proxies (YouTube views/likes/comments), IP/franchise labels and… | ||
| Optimizing Mobile Game LiveOps with Data-Driven Decisions | Lara Rajkovic | 30m |
Nanobit shares concrete LiveOps playbooks: RFM segmentation finds >35% of payers hibernating so targeted reactivation packs fail while a segmented seasonal… | ||
| Enhancing Personalization Using Recommendation Systems & Multi-Armed Bandits | Dmitriy M. | 27m |
Wargaming builds an offline recommendation pipeline that retrieves candidates with a lightweight recall model then re-ranks with a heavier ranker, uses… | ||
| Forget everything you think you know about NPCs and world-building | Constantin B. | 23m |
Claims that AI now makes feasible mass NPC creation — 10,000 or 50,000 characters — by shifting from per-NPC models to a shared, structured world knowledge… | ||
| Virtual Reality as a Practical Instrument | Alexey Shestakov | 28m |
Claims that VR cuts training time 40–60% reduction and boosts retention to 80% vs 20% come with concrete examples: an airline halves maintenance onboarding by… | ||
| Can Memes Predict Markets? Using Social Media Sentiment for Stock Forecasting | Milan Zdravkovic | 35m |
Claims that social media sentiment predicts prices get tested on mid‑liquidity Asian stocks (14 assets) and cryptocurrencies (≈13,000 hourly Reddit points)… | ||
| The Evolving Rules of Crypto and Blockchain | Emilija Vukovic, Milos Pavlovic, Milos Bosnjak | 42m |
Panel gives numbers and concrete shifts: illicit crypto flows total about $24 billion = 0.34% of volume (UN AML is $1.7T, so crypto under 1.5% of laundered… | ||
| Blockchain in Enterprise Systems: Integration of Blade Blockchain | Kosa Nenadic | 23m |
Blade Blockchain v1 integrates into AWS by federating enterprise IdPs into AWS IAM Identity Center, mapping each user to an on‑chain keypair stored in KMS HSM… | ||
| Advancements in Cryptography and Zero Knowledge Proofs - Binius Curve | Milica Vulic | 14m |
Claims that Binius Curve is an elliptic-curve-style construction implemented over binary fields to get native bitwise arithmetic, constant-time operations and… | ||
| AI Humans & Cross-Chain Metaverse: The Future of Data & Community Growth | Suren M. | 30m |
Claims a working cross‑chain metaverse that spans Solana, Avalanche, Bitcoin, Ethereum and Tron with CassandraAI avatars and a custom LLM pending patent… | ||
| Privacy-preserving Computation in Plain Sight | Aleksandar Veljkovic | 22m |
Describes concrete patterns for model training that let an observer verify results without seeing data by combining zero-knowledge proofs, fully homomorphic… | ||
| Blockchain Analytics with Dune | Mersiha Karamustafic | 24m |
Shows how Dune connects directly to Ethereum, Polygon, Optimism, BSC and Solana tables and runs SQL against on‑chain data to produce dashboards. Demonstrates… | ||
| Beyond the Balance Sheet: Transactional Data as a Window to Company Health | Jovan Milovanovic | 28m |
Claims that daily transaction data reveal company health faster than year‑end accounts and that Heartbeat turns that into actionable signals using heuristics… | ||
| AI Innovation for a Smarter Financial Ecosystem | Danka Jokic, Matija Ratkovic, Hans Kleinsman | 01h 03m |
Bank fintech runs a 25‑person data unit with 9 data scientists building sentiment and topic classifiers, a time‑series caller‑volume predictor, chatbots with… | ||
| AI in Financial Services: Automated Company Valuation | Goran Gvozden | 28m |
A web tool running on GCP automates valuation for six Zagreb Prime Market firms by extracting Excel and PDF financial statements into Postgres then converting… | ||
| The Fast and the Curious: Harnessing AI in Financial Data Extraction at Scale | Nikola Popov | 31m |
A production pipeline scrapes millions of scanned PDF filings, classifies pages, extracts 38 million fields across 3.6 million pages and validates results with… | ||
| Top Fraud Trends in Crypto for 2024 and How to Stay Ahead in 2025 | Ana Maria Anic | 24m |
Covers 2024 fraud vectors: synthetic identity, deepfake-enabled personalized phishing (created from social profiles and ChatGPT in minutes), insider risks and… | ||
| Data-Driven Approach to Process Automation in Banking | Dragan Vucic | 27m |
Argues that automation readiness depends on planned data use and not just tooling, showing three bank processes and one delivered AI-based OCR replacing an… | ||
| The Future of Global Taxation in the era of Gen AI | Vidak Rajovic | 22m |
Kogna shows a working pipeline that turns source ledgers into compliant GAAP and IFRS outputs using an AI-powered ETL plus a language model that generates… | ||
| AI has been achieved internally - The Journey | Miroslav Grbović | 57m |
Banca Intesa builds an internal AI-driven CRM: Python-based multiclass models produce next-best-product recommendations for 1.3 million customers and a… | ||
| Building Damage Detection from Satellite Imagery | Andrei Nita | 18m |
Claims damage detection from satellite and aerial imagery reaches about 80% average accuracy on hurricanes by combining high-resolution imagery (50–15 cm… | ||
| AI Solutions for Enhancing Customer Loyalty and Engagement | I.B., D.P., M.B. | 48m |
Panel covers deployed solutions and hard numbers: a fashion AI fashion assistant (5 weeks) ships and improves via customer feedback, a 3D avatar sizing (two… | ||
| Optimizing E-commerce: Data Science for Better CX & Sales | Youssef Kamal | 34m |
Shows RFM and cohort workflows with actual math: recency, frequency, monetary multiplied into a 12-point score example (12 months / 6 active = 60% consistency… | ||
| Transforming E-Commerce with LLMs: 3 Real-World Use Cases | Ratko Nikolic | 25m |
Boxalino demonstrates three deployed LLM systems: automated SEO and marketing content for PerfectHair.ch that claims up to 95% workflow automation… | ||
| AI Search for Marketplaces: Experiences & Tech | Mladen Fernezir | 31m |
Velebit AI presents deployed features: search by image using a custom CNN with binary vectors in production since 2016 for fashion, a recent multimodal vectors… | ||
| Demand Forecasting: Accuracy, Stability, Automation | Irena Bojarovska | 28m |
Market-level demand forecasts run for a 16-week horizon across 25 markets, produced daily and weekly with separate models for short and long horizons… | ||
| Data Driven Supply Chain Network Design | Mila Stankovic | 25m |
Pipeline that ingests SAP/WMS and external rental cost data into an SSAS data model with geocoding via API, then runs weighted k‑means (scikit‑learn) to find… | ||
| Machine Learning: The Engine Behind Modern E-commerce Search | Milutin Studen | 22m |
E‑commerce search moves from brittle keyword matching to embeddings and semantic retrieval that recover synonyms and context, with query classification into… | ||
| AI: Redefining Customer Experience | Darko Marjanovic, Vladimir Cuk | 24m |
Retailer case studies show recommendation engines moving conversion from 2% to 12% on 1 million members by consolidating logs, loyalty and browsing data… | ||
| AI in Retail - Advancing Promo Demand Forecasting | Bruno Segvic | 30m |
Claims a 2.5-fold reduction in forecasting error for a 400+ store rollout by combining LightGBM with feature engineering that models cannibalization… | ||
| AI-Powered Pricing: Smarter Decisions, Bigger Profits | Anthony Gale | 27m |
Claim: AI replaces guesswork with models that predict price elasticity and make real-time adjustments across SKUs, using competitor scraping, channel signals… | ||
| MMM and MTA: The Future of AI-Driven Marketing | Dobrica Cosic | 30m |
Explains how open source Robin MMM (Meta) and Markov models convert MMM and MTA into near‑real‑time engines: Robin spits out 15–20 allocation scenarios and one… | ||
| Scaling B2B Sales with AI Systems | Agostino Calamia | 21m |
Claims a working system that uses LLMs plus predictive ML to send AI-generated home cards and automated emails, ranking targets with propensity and uplift… | ||
| Future of AI | Alin Kalam, Ryan Bolick, Fabro Steibel | 38m |
Combines static analysis (using tree-sitter) with generative models to auto-document codebases, noting evaluation and productionization as the hardest step; a… | ||
| How is international collaboration on AI shaping sustainable development? | H.E. Anke Konrad, H.E. Luca Gori, H.E. Charlotte Sammelin, H.E. Pierre Cochard | 52m |
National plans and concrete projects: Germany updates its 2018 AI strategy and a 2021 National Data Strategy with €4–5 billion until 2025, ministries build… | ||
| How is international collaboration on AI shaping sustainable development? | Emanuele Giaufret, Dominic Otway, Yakub Beris | 23m |
UNDP and partners report concrete outputs: a public data portal Data.go.rs hosting 2,200 datasets from 110+ institutions, an operational supercomputer platform… | ||
| To risk or not to risk? The "Brussels effect" on AI regulations in Latin America | Fabro S. | 27m |
A rundown of concrete claims and surprises: a paper found Hollywood scripts in training data triggering strikes over copyright; EU impact work guessed 10%… | ||
| An EV Boom or a Gridlock? A Data-Driven Approach to UAE's Green Energy Transition | J. L. | 26m |
A UAE-focused platform models whether the grid can absorb a projected EV surge while hitting 30% clean energy by 2030 and 50% by 2050, combining AI… | ||
| SilkRoad: A Multilingual Communication Platform | Xian Yang | 45m |
SilkRoad is a platform from Panag Laboratory that deploys a Chinese foundation LLM (PUNON mind) extended to supports 50+ languages and layered services for MT… | ||
| Guardian - fraud detection using graph analysis | Tatjana Pejcinovic, Petra Sinka | 24m |
Guardian ingests company and ownership records into Postgres then builds graphs in Arango or directly in Python, aggregates monthly transaction amounts and… | ||
| Vision-Language Models for Fraud Detection | Renat Khizbullin | 23m |
Fine tuned a Florence VLM on internal ID images to create a single “universal document expert” that replaces dozens of specialized networks, training on up to… | ||
| AI in Cybersecurity & Fraud Prevention | Marko Elazar, B. Veselinovic, Petar Bajovic, Zarko Kecic | 52m |
Cybersecurity stacks produce billions of records/day that get distilled by XDR, NGFW logs and SIEMs into alerts, and the national registry reports >1 billion… | ||
| The Risk of AI adoption and business implication of not using AI | Kristian Kamber | 22m |
SplxAI offers automated adversarial testing and compliance automation for production LLMs, claiming to have found a multimodal chatbot breach that caught 5,000… | ||
| People-Centric Security: From Awareness to Culture | Uros Arsenijevic | 13m |
Argues that traditional quarterly courses fail to change behavior and shows a working people‑centric system that replaces 3–4 hour sessions with 3 minutes per… | ||
| The High Stakes Security of AI | Brian Wagner | 41m |
AI systems rely on RLHF with tiny percentages of human scoring and that creates real biases — cheap scorers in Nigeria push words like delve into outputs — and… | ||
| AI As The New Hacker Frontier | Tom Van de Wiele | 40m |
Claim: attackers already use AI to cut reconnaissance and scaling costs, turning noisy hammers into precise scalpels. Examples include initial access brokers… | ||
| The Future of Cybersecurity in an AI-driven Era | Francesco Vigo | 39m |
Generates concrete takeaways: LLMs can classify phishing and summarize logs but suffer prompt injection and data-poisoning; tests show injected whitespace or… | ||
| Eye tracking in bilingual neurolinguistics | Susanna Makerova | 45m |
Eye-tracking with an Eyelink Portable Duo records silent and aloud reading from a corpus of Russian and polysynthetic Adyghe sentences for 50 adult bilinguals… | ||
| Memristive Devices In Neuromorphic Systems | Victor Erokhin | 31m |
Organic polyaniline volatile memristors are built with a solid Li-doped polymer electrolyte and show hysteresis, rectification and tunable conductivity… | ||
| Neuromorphic computing in spiking neuron-astrocyte networks | Susanna Gordleeva | 31m |
Models biological astrocyte calcium as third-order nonlinear dynamics coupled by diffusion to spiking neurons (Hodgkin‑Huxley style astrocyte equations… | ||
| What can artificial neural networks learn from biological neuromodulatory systems? | S. R. | 27m |
Presents a dual framework that builds detailed biophysical cortical microcircuit models and distills them into layered recurrent spiking networks; implements… | ||
| Human Brain vs AI Brain: Can We Design a Truly Human-like AI? | S. R., D. O., J. V. | 52m |
Panelists argue current models remain abstract and miss cellular diversity, noting a 0.3 mm3 rat fragment contains 31,000 neurons and 55 types that networks do… | ||
| Recap and Closing Remarks: Neuropunk revolution and energy consumption | Max Talanov | 20m |
Argues that spiking neuromorphic hardware can cut AI energy use by orders of magnitude, demonstrates a hardware model of a rat spinal cord using memristive… | ||
| Neuro-Diverse AI: Enhancing Creativity through Unconventional Strategies | Jordi V. | 39m |
Claim: weaving non-neurotypical reasoning (hyperfocus, synesthesia-like pattern linking) and morphological computing into generative AI produces more… | ||
| Neuromorphic Linguistics: Advancing Emotional Intelligence in Neuro-AI | Irina K. | 38m |
Claims neuromorphic linguistics needs a non‑discrete, graded model of emotion—emotional intelligence gradation (general, emotionogenic, neuromorphic… | ||
| AI & Neurobiology of Deception: How AI Helps Us Understand Ourselves | Evgenia A. | 29m |
AI-driven analysis combines eye-tracking and HRV with skin conductance, breathing and 200 derived features, processed with NeuroKit and simple k‑means… | ||
| Quantum machine learning with superconducting circuits | Danijela Škugor Markovic | 28m |
Experimental quantum neuromorphic system uses superconducting resonators on a single niobium chip cooled in a dilution refrigerator, reading photon numbers… | ||
| Memristor-based neuroelectronics for artificial and hybrid intelligence | Alexey M. | 24m |
Claims practical memristor prototypes that implement analog vector–matrix multiply in a 4-kbit crossbar, demonstrate precise multilevel resistive tuning and… | ||
| Fireside chat: Challenges with AI implementation in organizations | Bojan Jovic | 26m |
Argues that AI adoption requires visionary leadership, a solid data backbone, and measurable KPIs rather than vague pilots; cites a logistics platform (called… | ||
| Introduction to Data Visualisation | Josip Saban | 31m |
Argues cloud data platforms consume, transform, serve data and AI is a consumer, not part of the platform; claims most cloud migrations fix legacy and… | ||
| Fireside chat: Role of Leadership in Scaling AI across the Enterprise | Natali Delic | 47m |
Serbia’s data ecosystem shifts when 2019 AI strategy acts as a tipping point, but enterprises are still 5–10 years late on preparing data: silos, missing data… | ||
| AI in Legacy Industries: Navigating Transformation | Marija Matokanovic | 28m |
Argues that legacy manufacturers and finance and healthcare firms need a strategic AI transformation, not piecemeal tools, and gives a five‑level maturity… | ||
| How to effectively manage AI/ML projects | Ana Stojkovic Knezevic | 26m |
Project management approaches tailored to AI matter: use Data Driven Scrum (DDS) with sprint decisions driven by data not fixed calendars, keep an updated risk… | ||
| How Alice, intelligent assistant, learned to understand people with speech impairments | A Shapedko | 24m |
Yandex upgrades Alice with a July 2024 model based on YandexGPT and real-world retraining on speech-impaired voices, using Yandex Tasks to collect about 1… | ||
| Data-Driven Growth & Efficiency | Goran Pavlovic, Dmitrii Pshichenko, Tiago Peixoto | 45m |
Panelists claim measurable wins and clear constraints: 50% of digital projects show results within two years, companies build internal platforms and train… | ||
| AI Powering Aston Martin F1 & Digital Transformation | Nikola Vurdelja | 26m |
Aston Martin F1 and Oran Telecom use NetApp storage, BlueXP and NVIDIA to treat telemetry as mission‑critical data — cars stream about 100 megabits per car and… | ||
| The Art of Telling Stories with Data | Gulrez Khan | 38m |
Claims that storytelling triples audience attention by framing insights as narratives and not raw outputs while showing concrete before/after work: a dashboard… | ||
| Evaluating the Worth of Implementation: Insights from Process Mining Projects | N. Vasileva | 20m |
Shows how process mining plus event logs finds concrete gains: a retail warehouse with ~20,000 orders/month has 21–25% suboptimal routes, 15% caused by article… | ||
| Integrating AI into office software | Ali Haidar | 18m |
OnlyOffice embeds AI across editors with an AI tab features set that does summarization, shorten/expand/simplify text, image generation from text or image, SEO… | ||
| Yottaanswers.com – Delivering Billions of Smart Answers Efficiently | Borislav Agapiev | 22m |
Claims a proprietary index of 27 billion vectors plus a separate 2 billion cause‑and‑effect vector store that together produce traceable, non‑hallucinating… | ||
| Future of the future: creating an AI product beyond hype | Radovan Bacovic | 23m |
GitLab builds an AI assistant called GitLab Duo inside a tight timeline, dogfooding it across self-managed, dedicated and private-cloud installs to collect… | ||
| Navigating the AI Project Lifecycle | Luka Anicin | 27m |
A stepwise roadmap maps problem definition through data, modeling, deployment and long term maintenance with concrete checks: can the data be accessed and is… | ||
| Balancing Personalization & A/B Testing in a Customized World | Joy Chatterjee | 29m |
Compares classic A/B testing (example: Booking.com changed “Book now” to “I'll reserve” and measures conversion rates) with adaptive approaches that reallocate… | ||
| Future of work with google workspace, be ready to start the journey now | Suzana D. | 35m |
Google Workspace embeds Gemini to cut time‑to‑market from weeks to hours (team paper example created in one hour), claims a survey of 2,500 senior leaders… | ||
| Boosting Data Science Impact by Removing Bottlenecks | Mikheil Nadareishvili | 22m |
Claim: quadruple data science impact to eight figures in one year by removing process bottlenecks and changing structure. Team shrinks active projects from 60+… | ||
| Why AI Projects Fail: The Missing Link of a Strong Data Strategy | B. P., L. S. | 20m |
Claims that corporate AI projects hit an 80% failure rate and that only 4.7% of AI projects create real innovation, traced to five root causes: unclear problem… | ||
| Maximizing Data Usability for Impact | Filip Petrovski | 33m |
Companies spend 90% data preparation 10% analysis and the goal is to flip that to about 10% prep 40–50% analysis by building a business-facing Discovery layer… | ||
| Data & Analytics Strategy: 5 Overlooked Components | Alex Barakov | 26m |
A concrete playbook for stopping the cycle of low adoption: a prioritization framework that scores impact versus effort, a maturity model with… | ||
| Alstom’s Journey to Safe, Reliable AI Mobility | Nenad Mijatovic | 29m |
Alstom deploys AI across non‑safety systems (about 20% of business) with a 250‑person global team, reusing camera, odometry and signaling data to build… | ||
| Leveraging Data to Unlock Business Value | Roberto De Meo | 38m |
Visa shows concrete results using transaction and merchant data to boost revenue and cut losses: a fraud detection system doubles fraud detection (from ~25% to… | ||
| Exploring LLMs across clouds – A Year in the Life | Tomislav Tipuric | 31m |
Side-by-side comparison of cloud LLM stacks: Google’s Gemini 1.5 offers multimodal strength and charges about $1.75 input / $5 output per million tokens… | ||
| How to successfully scale GenAI in big organizations | A. A., J. H. | 47m |
Deloitte describes moving GenAI from PoC to production using a 2,000 C‑level survey that finds ~40% of firms feel partly ready but only 5–6% fully ready and… | ||
| Building an AI Platform | Livio Bencik | 25m |
Builder describes a six‑year journey that launched dozens of AI products before generative AI, starting from half a dozen disparate projects and an internal… | ||
| Scalable and Sustainable AI Infrastructure (On The Seabed) | Peter Stensgaard Hansen | 26m |
Europe's data centres use over 150 TWh a year and spend ~40% on cooling (equal to CO2 from 17 million households), so the plan places prefabricated 20 ft units… | ||
| Journey to AI driven company | Azfar Shah | 26m |
Company becomes AI-driven over a 3‑year journey, creating an in‑house AI unit of data engineers, software engineers, data scientists and stewards that reports… | ||
| BI Battle: Centralized Control vs. Decentralized Agility | Stefan Stosic | 18m |
Presents four BI organization types—full centralization, hybrid centralization, hybrid decentralization, and full decentralization—and evaluates them against… | ||
| The “Secret Sauce” of Canada’s Top Public Sector Data Science Team | Kris Andreychuk | 24m |
Edmonton centralizes data science since 2021 and claims a four‑ingredient "secret sauce": a well‑defined problem, a willing stakeholder, clarity on who or what… | ||
| Building a Data-Driven Culture: What the C-Suite Needs to Understand First | F. Mikulicic | 28m |
Argues that culture is an effect, not a lever, and the real intervention is the company’s nervous system: decision flows, control points and integration gaps… | ||
| Importance of Data Maturity before embracing AI | Dejan Djekic | 25m |
Claims that AI succeeds only after investing in data accessibility, data integration and governance, backed by concrete infrastructure steps: ingestion, block… | ||
| Strengthening Security and Compliance in Our Data Platform | Dennis Van Rooijen | 17m |
gitlab hardens its data platform by hiring a data security engineer, automating CI security scans (SAST, DAST, secret and dependency checks) and running… | ||
| Federated Data Architecture | Miodrag Pesovic | 31m |
UniCredit Bank Serbia adopts a modified Data Mesh: a source-first model that builds one source per source-aligned data product and composes consumer-facing… | ||
| TIS Group’s AI Journey: From Healthcare to Agriculture | Tomislav Strgar | 25m |
TIS.AI runs a commercialized, EU‑co‑financed sendd.eu infant movement detection product that uses web/mobile apps, videos labeled independently by two… | ||
| How AI Agents Will Change the Future of Programming | Dmitry Krasnov | 30m |
Claims that agentic code assistants use tools to open and edit files, run unit tests, and search the web and follow a three-step loop of planning then… | ||
| Building next-gen speech synthesis for Bookmate audiobook service | Vladimir Platonov | 22m |
Baseline TTS trains on 90 hours of training data and deploys across more than 100,000 books; voice selection narrows ~50 candidates to two final voices… | ||
| AI Development and Copyright: Public ≠ Free Use | A. R., B. K., D. M. | 50m |
Panel argues that public data is not free: training on scraped YouTube or web content carries copyright risk, and models trained without clear licences can… | ||
| Optimizing AI Inference with ML Compilers & Hardware | Milan Stankic | 34m |
Argument: ML compilers (TVM, MLIR, Triton) and hardware decisions determine whether LLMs are practical or too expensive — estimates show up to 90% of total… | ||
| Hominis: Training Human-centric Large Language Models | Narendra Patwardhan | 22m |
Hominis is a 13B model trained to match or beat Llama 2 while claiming 50x less carbon; training ran eight months for pretraining and four months for… | ||
| The Double-Edged AI Sword: Examining Applications in Different Fields | Sead Delalic | 27m |
Claims and concrete takeaways: more than 40+ production AI solutions were built across transportation, supply chain and healthcare; a Microsoft chatbot… | ||
| Beyond Supervised Learning with Zero-Shot Vision | Nemanja Milosevic | 28m |
Zero-shot classification uses a joint image–text embedding space to label unseen classes by similarity rather than softmax, trained with contrastive learning… | ||
| Fostering Trust in AI-Driven Finance | Orsalia Andreou | 02h 01m |
Argues that trust must be built across the entire model lifecycle, from business case and stakeholder mapping to data, feature engineering, model discovery… | ||
| Building a Data Experimentation Platform Without a Platform Team | Vladislav Metelev | 31m |
Builds a full experimentation stack built twice in five years without a platform team, combining client-side feature toggles, caching and event tracking with a… | ||
| 2025: year of Ai dilemma - ethics, regulations and innovations | Sray Agarwal | 24m |
Argues irresponsible AI creates three tangible costs—revenue, reputational, regulatory—and cites a survey claim that responsible AI projects earn about twice… | ||
| Application of AI technologies in nuclear power plants | Peter Kertys, Maros Buban | 21m |
An internal LLM chatbot is built with Python backend, JavaScript front end, Azure AI Search vector indexing and SQL lookups, using a multi‑agent pipeline and… | ||
| Path of Data & AI Challenges | Ivana Sesis, Anja Beric, Dejan Grubisic | 46m |
Radiology workflows move from paper archives to one‑click digital access, enabling AI to cut reporting backlog but raising patient privacy and data curation… | ||
| Open Data for All: Social Action, Peace Tech & Community Outreach | Nathan Coyle | 26m |
Argues that open data can power grassroots peacebuilding but is blocked by corporate designs and siloed government processes, offering concrete tools: Skild… | ||
| Time Series Forecasting & Anomaly Detection in Radio Networks | Jelena Tomic, Ivana Jovovic | 31m |
Network telemetry arrives at scale — roughly 3,000 KPIs per cell per hour, 3k–60k cells per country and ~150 KPIs per call — and forecasting feeds an adaptive… | ||
| AI Trustworthiness: Path to Mass Adoption in Critical Industries | Arnault Ioualalen | 31m |
Claims that standards can make high‑risk AI certifiable by 2025 are backed with concrete work: ISO already has 30 published and 36 ongoing standards and aims… | ||
| Insights & Engagements with Customer Data Platforms | Pratul Kumar Chakravarty | 31m |
Argues that Customer Data Platforms plus enterprise data hubs deliver real, measurable outcomes: claims include CAGR 40% to $28B by 2028, model-driven… | ||
| Effective Data Migration: Challenges & Success | Marcin Szymaniuk | 26m |
Practical guidance for migrating data platforms to public or hybrid clouds with concrete tradeoffs and fixes: choose tools to match team skills, build… | ||
| How gender imbalance is holding back the future of AI | Jovana P., Olga M. M., Slobodan M. | 47m |
Identifies concrete failures: voice assistants default to female personas, Google Translate flips Serbian pronouns, facial recognition and health AIs lack… | ||
| Building up the Bosch Semantic Data Lake | Sofija Pervulov | 29m |
Bosch builds a production Semantic Data Lake that layers a scalable data lake with a knowledge layer, ingesting batch and stream, structured and unstructured… | ||
| Predicting Retention on the BBC Sounds Service Using Historic User Dynamics | R Timms | 28m |
BBC Sounds uses temporal features and static demographics to score the probability each user returns next week, training on 100,000 users with 140 days of… | ||
| Modern Web Data Extraction: Tools, Techniques & Ethics | Domagoj Marić | 32m |
Modern web data extraction supplies training and monitoring data for generative and traditional AI but runs into a legal gray zone — LinkedIn/HiQ legal risk… | ||
| Sustainable & Smart AI Factories in Europe | Andrew van der Haar, Thomas Schmidt, Gauvain Girault | 51m |
Panelists call the €100B CERN for AI idea ambitious but under-specified, noting concrete bottlenecks: supply lag for B200 GPUs, spine‑leaf networking changes… | ||
| The State of Data Engineering | Jesse Anderson | 31m |
Claims that the industry sees an 85% failure rate for big data projects and declining average value creation over time drive a diagnostic: parallels with… | ||
| Scalable AI & Web Insights with GitLab, Snowplow & Snowpipe | Ved Prakash | 28m |
GitLab builds a CI/CD-managed pipeline using Terraform, Snowplow (collector & enricher), S3 staging, Snowpipe for continuous loads into Snowflake, DBT… | ||
| Turn Data into Action: How visualization aids decision intelligence | Sumeyye Suri | 21m |
Argues that clear visuals change outcomes by lowering cognitive load and speeding decisions, showing a concrete case where separate error histograms hide… | ||
| Harnessing Power: Inside Sofascore’s Data-Driven Infrastructure | Karlo Knezevic | 30m |
SofaScore runs a hybrid pipeline using Firebase -> BigQuery + NATS -> ClickHouse with JetStream, Kubernetes and Amazon S3 backups, claiming support for ~25M… | ||
| AI-driven query plan optimization in big data processing tools | Vadim Opolski | 28m |
Claims AI models trained on query telemetry pick Spark configs (e.g. spark.sql.cbo, join reorder, star-schema detection) that cut execution time often 2–3x and… | ||
| Data Platform Infrastructure at Rivian: A Deep Dive | Veljko Tanjga | 30m |
Rivian describes a production data platform that ingests vehicle telemetry via Kafka, processes streams with Flink, and serves time-series queries from Druid… | ||
| Event Analytics - from unstructured to insightful | Boris Paunovic, Igor Djordjevic | 24m |
Builds a subscription state machine that reconciles messy multi-source events by inserting implicit events (e.g., add a missing start or resume timestamp “a… | ||
| Transforming Automotive Infotainment with Data | Nemanja Tiosavljevic | 28m |
Rivian builds its infotainment analytics stack from scratch and collects de‑identified events from a roughly 90k vehicles fleet via a proprietary… | ||
| The (mis)use of GenAI in Data Analytics | Guilherme Diaz-Berrio | 25m |
Generative models often get misapplied to forecasting and regression where simple ETS or ARIMA beat transformers, and using LLMs for time series or sales… | ||
| Data Analysis in the Age of Data Democratization | Ioan Enache | 22m |
A data marketing agency outlines a practical pipeline: inventorying available sources, adding partner extracts (Spotify and an in‑house transaction scraper)… | ||
| Business Decision-Making in the Era of AI | Benjamin Dzubur | 26m |
AI already automates connection, cleaning, modeling and dashboarding and can produce a full interactive dashboard in 20 seconds from a CSV; demo uses Zebra BI… | ||
| Beyond the Hype: Realities of AI in Business | Martina Naumovska Necheska | 26m |
AI is not plug-and-play: a state‑of‑the‑art NLP model works for one customer but fails to scale when data sits in silos, so businesses must fix four… | ||
| AI-Powered Insights Mining:CBS BrandPulse Solves AI Hallucinations | Milan Bjeloglav | 23m |
CBS BrandPulse integrates with OpenAI by supplying a knowledge DB that describes metadata and by handing the LLM pre-cooked numbers and relevant statistics… | ||
| Alisa AI: Future-Proofing Your Business with Demand Forecasting | Nino Pozar | 33m |
BE-terna’s Alisa AI claims to replace manual planning by connecting to ERPs and running ML forecasts at the SKU level with features for promotions… | ||
| Data Driven Culture in Software Defined Vehicles | Erhun Arkan | 26m |
Rivian describes concrete systems: a Rivian Charge Score launched six months ago that uses fleet telematics to rate chargers and feeds trip planning, producing… | ||
| CI/CD Delays: Losing Time, Money & Sanity | Darija Strmecki, Lamia Cero | 25m |
Claims that developers lose 3–5 hours/week waiting and that up to 70% report increased stress from bad CI/CD. Describes an AWS implementation using change sets… | ||
| Encouraging Kids to Read: LLMs for Education & Fun | Ewelina Kucal, Maciej Dziezyc | 30m |
SoftwareOne builds the Book Aligned Activities AI Engine for Worldreader using AWS (S3, Lambda, Step Functions, Bedrock) and a Claude-style model to extract… | ||
| AI Agents: Transforming Autonomous Decision-Making | Lana Malic | 34m |
Claim: chaining tools and planner logic on the same LLM boosts zero‑shot code success from 45% to over 85%; implementations rely on OpenAI models plus… | ||
| How to fail AI implementations | Matevz Cerne | 23m |
Three concrete failure modes: poor scope (overpromised AI when a short user-focused pilot was needed), data prep (messy Excel/PDFs and inconsistent labels make… | ||
| Beyond Autocomplete: Local AI Code Completion Demystified | Daniel Savenkov | 29m |
JetBrains ships a local AI code completion that runs a 100M-parameter Llama model (1500-token context) quantized to 4 bits occupying ~50–500MB and using lama… | ||
| Accelerating AI with Hyperscaler Platforms | Sinisa Arsic | 31m |
Telekom Serbia migrates its analytics stack from IBM Cloud Era to AWS, using AWS SageMaker, EC2 and Lambda to handle a 300 TB monthly signaling feed and train… | ||
| Rivian: Scaling Data & ML Platforms | Aleksandar Cvejic | 24m |
Rivian describes building Polaris data platform to manage ADAS data that grows from 6B to 2T records/day and from 14 TB to 1.6 PB in six months, achieving… | ||
| DSPy: The End of Prompt Engineering as We Know It | Boris Cergol | 31m |
Open-source framework DSPy (short for declarative self-improving python) replaces manual prompt crafting with a declarative program of signatures, reusable… | ||
| Accelerating Model Deployment with MLOps | Kevin Dietz | 23m |
Claims that CI plus MLOps cut the 30+ day bottleneck for deploying models by automating the pipeline: a reproducible demo uses a repo with Dockerfile and… | ||
| Optimizing Context Tuning for RAG Systems | Sofia Konchakova | 28m |
Shows concrete outcomes: a data pipeline with automated updates, monitoring and chunking experiments that compare fixed-size chunking versus recursive… | ||
| VerifAI: Biomedical QA with Verifiable Answers | Nikola Milosevic | 31m |
VerifAI is an open-source biomedical QA pipeline that combines lexical OpenSearch and semantic Qdrant retrieval over ~35 million PubMed abstracts, uses a… | ||
| Boosting Sales with Intelligent BI Agents | Simun Sunjic, Lovro Matosevic | 30m |
A BI system is presented that turns natural language into validated SQL using Llama 3.1 fine-tuned models, Retrieval-Augmented Generation and vector… | ||
| Where there's a will, there's a way: The mad genius of LLMs as a classifiers | K. Munro | 30m |
Argues that LLMs can replace or augment intent classifiers in production if used selectively, with Swisscom using 1,300 labeled utterances to prototype… | ||
| All About Biases in Artificial Intelligence | Karen de Sousa Pesse | 41m |
Showcased are concrete harms and fixes: social media ads shift after changing gender, a chatbot turns racist in four hours, hiring models downrank CVs… | ||
| Securing Executive Buy-In for Data Strategies | Amie Bright | 24m |
Five concrete steps for turning data into measurable business value: engage cross‑functional teams, set clear outcomes, measure performance, communicate… | ||
| Fighting AI with AI | Rana Gujral | 47m |
Shows that deepfakes are already practical—one audio clone takes 72 hours to create on powerful GPUs and fools 90% of people, so defenses must also use AI… | ||
| Intelligent Interfaces and operating systems | Igor Sevo | 49m |
Claims a working toy hybrid OS that runs LLM prompts alongside deterministic interpreters (C, Python) using a ZAML format to store a global prototype memory… | ||
| AI in International Trade: Successes & Challenges | Peter Piehler | 24m |
Global trade is framed as 80–90% classification problem where AI automates routine tariff, export-control and sanctions decisions but struggles with the ~20%… | ||
| Overcoming Loneliness with LLM Dating Assistant | Dani El-Ayyass | 22m |
Social Discovery Group builds a dating copilot that generates two suggested replies in-app, trains models by fine‑tuning frontier APIs, and deploys across a… | ||
| Revolutionizing Access to Global Knowledge | Miodrag Vladic | 25m |
Claims a proprietary YottaAnswers Smart Answer Engine built from scratch that indexes roughly 30 billion answers and runs a lightweight Source Frontier… | ||
| Manufacturing in 21st century: AI supported manual assembly | Krste Pangovski | 37m |
Claims that human operators still do most manufacturing supported by data: over 120 million manual assemblers and a $17 trillion sector losing about $3… | ||
| Structuring Data for Scalable AI Applications | Mikhail Rozhkov | 29m |
Open-source Data Chain turns terabytes of unstructured files into a small SQL table of pointers and metadata, versioned and queried via a Python API that maps… | ||
| Yandex Alice: music control with gestures | Darya Vinogradova | 19m |
Built a gesture music controller that beats Google MediaPipe on precision and range by combining a hand detector and a separate lightweight classifier… | ||
| AI-Powered Real-Time Quality Control | Donovan Spronk, Vladislav Belov | 36m |
An AI vision pipeline deployed in Docker on edge devices and Azure integrates with PLCs and IoT hubs, using Kean/Cognex cameras, background‑removal plus… | ||
| Key lessons from building Data and AI systems over the last 26 years… | Dragan Tomic | 43m |
26 years of systems work distills into concrete claims: built core pieces of SQL Server crown jewel and re-engineered OLTP for cloud; moved to Databricks and… | ||
| The tech giants in AI - who is going to win the battle? | Ido Engel | 46m |
High-level market map with concrete bets and costs: Nvidia pivots beyond GPUs with $300 million and $700 million acquisitions to own performance and… | ||
| How Biometrics Are Shaping the Future | Dmitry Markov | 24m |
Vision Labs describes a deployed suite of face, body and gesture recognition models used in banking and safe-city systems, claiming 99.8% accuracy even with… | ||
| Multimodal Video Face Liveness: A Better Alternative? | Efim Boieru | 23m |
Argues that combining modalities beats single-frame selfies: a multimodal pipeline uses RGB + depth embeddings (depth collected from an iPhone TrueDepth sensor… | ||
| Building High-Performance, Sustainable AI Clusters | Thomas Kitzler | 25m |
Claims a jump from Titan (2012) to GB200 systems (2024) with roughly 800× energy efficiency and a cost drop from $130M to $400k for comparable work, cites a… | ||
| AI in maritime - Autonomous ships | Marina Ivasic Kos | 24m |
Camera footage from an LNG tanker and 14,000 annotated images (plus synthetic augmentations) train sea‑state classifiers and object detectors; a visual… | ||
| Will an LLM replace me? A Systematic study of Gen AI's impact on data jobs | F. Soomro | 24m |
Systematic review of ~40 Argusa projects by 11 consultants finds LLMs speed debugging, write documentation and summarize meetings but do not shorten a 14-week… | ||
| How Intelligent Compute Accelerators (ICAs) are Shaping the Future of HPC & AI | J.P | 47m |
Links text and image spaces using CLIP 768‑d vectors trained on roughly 400M alt-tag pairs, and generates images by a denoising diffusion process implemented… | ||
| Boosting Data Efficiency with AI: Insights from a Case Study | Christian Casari | 21m |
Sunny Cars implements email automation, RAG-powered knowledge management and LLM-agnostic pipelines (on‑prem Llama or cloud GPT‑4) to classify nine post‑rental… | ||
| Exploring the Mechanisms of LLM Reasoning Capabilities | Arsenii Shkunkov | 27m |
Presentation argues that LLMs often produce the appearance of reasoning but mainly rely on pattern matching and memorization, not robust logical inference… | ||
| Multi-Agent AI: Transforming Business Operations and Value Creation | Radu Orghidan | 27m |
Argues that autonomous multi-agent systems move beyond chatbots by doing intent analysis, planning and execution, and shows two deployed solutions: an AI… | ||
| Leveraging multimodal language model for AI native wireless systems | Ali Mokh | 33m |
Claims a working proof-of-concept that uses multimodal generative models to make wireless networks “AI native,” implementing a three-agent pipeline: a small… | ||
| How to train a large diffusion model | Eugene Lyapustin | 30m |
Built a Yandex image generator trained to predict compressed latents rather than pixels using VAE latents, conditioned on CLIP/T5-style text embeddings and… | ||
| LLMs - application and impact on business | Sacha Khanin | 40m |
Company builds hardware‑agnostic visual‑language models that run without GPUs for national projects in low‑resource languages, collecting spoken and visual… | ||
| Fine-tune LLMs from Kaggle Models using (Q)LoRA | Gabriel Preda | 32m |
Shows hands-on fine-tuning inside Kaggle Models notebooks using quantized Low-Rank Adaptation: load a model, apply 4-bit quantization, insert LoRA layers and… | ||
| Hyper-Personalized Marketing with Digital Platforms & Generative AI | Darije Ramljak | 26m |
An end-to-end pilot builds a custom Marketer Workbench on Salesforce Marketing Cloud that pulls customer data from Databricks on Azure, runs an AI… | ||
| 7 Deadly Sins of AI (And How to Avoid Them) | Tobias Zwingmann | 46m |
In his talk, Tobias revealed the seven most costly AI pitfalls that have derailed even the best companies. He highlighted how AI is transforming businesses… | ||
| nothing matches | ||