AI_dev: Open Source GenAI & ML Summit Europe 2024
2024-06-19..2024-06-20
Paris, France
as the organisers put it
Register sponsor SUBMIT to speak View the Schedule See Who’s Attending how to access live stream Post Event Report Thank you to all who joined us for AI_dev: Open Source GenAI & ML Summit Europe 2024!
| title | author | length |
|---|---|---|
| AI_dev: Open Source GenAI & ML Summit Europe 2024 Highlights | 02m | |
Check out all the session recordings here: https://www.youtube.com/playlist?list=PLbzoR-pLrL6oII-dLqw9vpif0U_kTycj6
Stay up to date on all things AI, don't… | ||
| Keynote: Welcome Back | Ibrahim Haddad | 01m |
LF AI & Data Foundation sets priorities for Europe focused on scaling open source GenAI and ML projects with concrete goals: create shared tooling for… | ||
| Accelerating the LLM Lifecycle on the Cloud | Luca Antiga | 33m |
Lightning AI presents an open source stack (PyTorch Lightning, Fabric, TorchMetrics, litgpt, litdata, litserve, Thunder) that standardizes training and serving… | ||
| Hallucination-Free LLMs: Strategies for Monitoring and Mitigation | Wojtek Kuberski | 39m |
Describes practical monitoring and mitigation: two main families — LLM self-evaluation (use one LLM to score answers, e.g. SelfCheckGPT / LLM-eval) which needs… | ||
| From OpenAI to Opensource AI: Navigating Between Commercial Ownership and Coll... | Raphaël Semeteys | 39m |
Compares actual openness across weights and datasets and scores model artifacts with a 0–4 openness matrix (0 closed, 4 fully reusable). Calls out… | ||
| Powering Your Generative AI Workloads with AMD and Open-Source ROCm | Farshad Ghodsian | 32m |
AMD hardware plus open-source ROCm runs real generative workloads: PyTorch works on Linux, Optimum-AMD and Transformers run via Llama.cpp/AMA, and… | ||
| Designing AI for Network Troubleshooting: A Langchain Proof of Concept | Jesus Illescas | 30m |
Proof of concept ties a webhooked Grafana alert to a FastAPI service that spins up Langchain agents which call Python tools that use a PATS connector to query… | ||
| Wikidata Knowledge Graph to Enable Equitable and Validated Ge... | Jonathan Fraine, Lydia Pintscher | 31m |
Wikidata provides an open CC0 knowledge graph of over 100M crowd-sourced items across ~300 languages and is being vectorised so LLMs can use it for retrieval… | ||
| Introduction to Distributed ML Workloads with Ray on Kubernetes | Abdel Sghiouar | 26m |
Introduces Ray core concepts (actors, tasks) and shows running Ray on Kubernetes using the KubeRay operator, translating Ray jobs into Kubernetes CRDs and… | ||
| Between Public Benchmarks and Business Needs | Lukasz Borchmann | 24m |
Open-source models and recipes are available: Arctic llm, Arctic tilt and Arctic embed with Apache 2.0 weights, code and a published cookbook. Arctic llm uses… | ||
| Building Reproducible ML Processes with an Open Source Stack | Einat Orr | 33m |
Claims a reproducible ML experiment needs three concrete pieces: code (Git + KubeFlow), data (MinIO + lakeFS) and environment (infrastructure-as-code), and… | ||
| Advancing Responsible AI: Unveiling the Software Carbon Efficiency Rat... | Chris Xie, Tereze Gaile | 35m |
Green Software Foundation's SCER (aka Shear) defines a reproducible four-part rating: categorization, benchmarking, rating, labeling, and requires CO2e metrics… | ||
| The Dark Side of AI: The Hidden Supply Chain Risks in Open... | Jossef Harush Kadouri, Tzachi Zornshtain | 45m |
Demonstrates concrete supply‑chain attacks: account takeovers of popular npm packages with millions of weekly downloads, typosquatting that installs… | ||
| Examining the Principles of Observability and Its Relevance in LLM... | Guangya Liu, Jean Detoeuf | 19m |
Describes how an OpenTelemetry collector plus Traceloop instrumentations capture per-query fields like vendor prompt temperature accuracy model_version and… | ||
| Workshop: Deploy and Monitor ML Pipelines with Open Source and Free Applications | Rami Krispin | 45m |
Builds a reproducible pipeline that pulls US hourly electricity demand from the EIA API, handles the API 5,000‑row limit with pagination, and backfills six… | ||
| Fireside Chat: Harnessing AI for Financial Services Innovation | Luca Borella, Jane Gavronsky | 41m |
FINOS runs an AI Strategic Initiative with three tracks: AI Readiness, LLM Exploration and a research fund that pays for benchmarks and data work (not… | ||
| Towards a Practical Ethics of Generative AI | Geert Hofman | 32m |
Research project builds a usable ethics layer for creatives by keeping the Double Diamond and injecting measurable checkpoints: a proposed openness score… | ||
| Building a Data Infrastructure for AI/ML | Keith Pijanowski | 31m |
Claims that Open Table Formats Apache Iceberg Apache Hudi Delta Lake make object storage a full data warehouse and show a reference architecture that supports… | ||
| Code Security Reinvented: Navigating the Era of AI | Joseph Katsioloudes | 32m |
GitHub Security Lab reports 1,000+ issues found in open source with 600+ CVEs and a 96% fix rate; demos show using GitHub Copilot to write safer code, run… | ||
| AI and Incident Management: How to Reduce Manual Work and Speed up the Incident... | Birol Yildiz | 30m |
ilert GmbH builds a SaaS incident platform that demos an OpenAI function calling conversational UI to produce structured on‑call schedules (bi‑weekly demo… | ||
| What's in a Vector Database? | JP Hwang | 39m |
Walks through building a usable vector database from an object store to indexes to sharding and replication, demonstrating that a working prototype is possible… | ||
| Panel Discussion: Navigating the Ethical Landscape: Responsible... | Anni Lai, Mirko Boehm, Oita Coleman, Pedro Ortiz Suarez, Adrian Gonzalez-Sanchez | 49m |
Generative AI Commons reports over 100 active members across five workstreams running bi‑weekly meetings and is producing a responsible AI framework first… | ||
| Keynote: Welcome & Opening Remarks | Ibrahim Haddad | 12m |
Opening housekeeping and a short set of announcements about LF AI & Data Foundation activity, noting an initiative launched in September 2023 that aggregates… | ||
| Moxin: The All-In-One Tool to Run LLMs on Your Personal Device | Michael Yuan | 39m |
Moxin is an open source desktop app that runs LLMs locally with a Rust-based GUI and a backend using the WasmEdge runtime. Demo shows it running an 8B Llama 3… | ||
| Algorithmic Alchemy: Transforming Drug Discovery with Open AI | Gloria Macia | 36m |
Argues that AlphaFold-style models truly change preclinical work by predicting 3D protein structures with near-experimental accuracy — GDT score 92.4 — but… | ||
| Applying Open Source Methods to Building and Training Large Langu... | Carol Chen, Mark Sturdevant | 30m |
Project creates a workflow that lets contributors add 5-example YAML skills or document-sourced knowledge to an upstream model without full fork or fine-tune… | ||
| RAG Pipeline Evaluation | Estelle Scifo | 40m |
Built an evaluation pipeline that runs Neo4j-backed RAG pipelines against a dataset of 10,000 movies and 20,000 persons, compares three retriever strategies —… | ||
| AI Software Bill of Materials: From the EU AI Act to Your MLOps | Adrian Gonzalez Sanchez | 32m |
Proposes an AI Software Bill of Materials: a JSON spec that encodes EU AI Act transparency fields (provider, market, data types, high‑risk flags) and lives in… | ||
| Introducing the Model Openness Framework: Achieving... | Anni Lai, Matt White, Cailean Osborne | 26m |
Model Openness Framework defines a practical checklist for deep‑learning models with 17 components (code, data, documentation, weights), recommends treating… | ||
| Workshop: Working with Gemma and Open LLMs on Google Kubernetes... | Abdel Sghiouar, Victor Dantas | 01h 15m |
Hands‑on workshop that provisions a RAG app via Terraform on Google Kubernetes Engine, fine‑tunes and serves Gemma fine-tunable models pulled from Hugging Face… | ||
| Rethinking Software and Software Engineering in the Era of Foundation Models... | Ahmed E. Hassan | 33m |
Claims ten concrete pain points labeled 10 SE4FMware challenges, cites operational costs where a regression test suite can add $3–$5 per CI checkin, and shows… | ||
| Keynote: DBRX: Training Modern LLMs on Databricks | Niall Turbitt | 16m |
Demonstrates concrete LLM training pipelines on Databricks that combine Delta Lake for curated training data, MLflow for experiment and model tracking, and… | ||
| Keynote: The Common Pathway of AI and Software | Waren Long | 15m |
Argues that treating AI as an engineering pipeline produces pragmatic software results: a tested chain of planning module, knowledge retriever, and code… | ||
| Workshop: Optimizing Kubernetes Cluster Scaling for Adv... | Shivay Lamba, Shivanshu Raj Shrivastava | 01h 22m |
Shows a runnable path: Terraform + Civo spins a GPU-backed Kubernetes cluster in ~10 minutes and deploys Olama + OpenWebUI hosting Llama‑3 (demo succeeds… | ||
| Future-Proofing Agent Supervision | Alexandre Variengien, Diego Dorn | 31m |
EffiSciences' BELLs project curates a GitHub corpus of labeled traces (jailbreaks, prompt injections, hallucinations), keeps some classes and examples private… | ||
| Workshop: Choosing the Best Open Source LLM for Your Application | Nikolai Liubimov | 59m |
Claims that leaderboards mislead and human supervision is essential for real-world LLM selection; benchmarking shows about one in six benchmark queries… | ||
| Lightning Talk: Open Source GenAI: A Community-Powered Future | Ofer Hermoni | 10m |
Organization created in 2018 now hosts 70+ projects with contributions from over 3,000 organizations, ~100,000 contributors (30,000 active), ~200M lines of… | ||
| Toward AI Democratization with Digital Public Goods | Lea Gimpel, Daniel Brumund | 43m |
Sees concrete commitments: 70% of SDGs rely on digital solutions; the DPG standard uses 9 DPG indicators (including CO2 footprint and responsible AI… | ||
| Workshop: Efficient and Portable AI / LLM Inference on the Edge Cloud | Xiaowei Hu | 48m |
Shows a working pipeline that swaps heavy containers for a tiny Wasm binary and runtime, with wasm runtime <20MB vs PyTorch images ~3GB and C++ toolchains… | ||
| Keynote: Navigating Turbulent Waters of AI: Building a Secure and Safe Foundation | Dan Lorenc | 16m |
Generative AI multiplies code volume and attack surface and attackers exploit vulnerabilities faster than defenders, so the claim is that teams must harden the… | ||
| Keynote: New Advances for Cross-platform AI Applications in Docker | Michael Yuan, Justin Cormack | 12m |
Docker adds Docker support for WebGPU so container apps call host GPUs and accelerators through a single portable API, removing the need for separate… | ||
| Keynote: Open Source AI Everywhere | Omar Sanseviero | 16m |
Overview of the open source AI ecosystem, claiming hundreds of large models available and a push to let anyone fork and modify weights and pipelines; shows… | ||
| The Open Source AI Dilemma: Crafting a Clear Definition for Open Source AI | Ofer Hermoni | 37m |
Linux Foundation AI shows 67 projects, >100,000 contributors (≈30,000 active) and 250,000+ GitHub stars and launches a concrete model openness framework that… | ||
| OpenSearch: Conversations with Context | Lucas Jeanniot | 26m |
Shows how OpenSearch links vector indexes to LLMs (OpenAI, Anthropic, and local LLaMA3 runs) to do retrieval augmented generation with a memory index that… | ||
| Meet DBRX, the New Standard for High-quality LLMs, and How... | Anastasia Prokaieva, Niall Turbitt | 41m |
Databricks unveils DBRX as an open source mixture-of-experts LLM claiming 132B parameters with 36B active and up to 3x faster inference; it trains on 12… | ||
| A Hopeful View of Truly Open Source AI | Julia Ferraioli, Tom Callaway | 39m |
Argues that full transparency — code, model weights, provenance and training data — is necessary so organizations can audit bias, verify provenance and reduce… | ||
| Efficient and Cross-Platform LLM Inference in the Heterogenous Cloud | Michael Yuan | 33m |
WasmEdge implements WASI-NN so Rust apps compile to one Wasm binary that runs unchanged on Nvidia CUDA servers, ARM NPUs or Mac Metal and is deployable with… | ||
| Cloud Native Artificial Intelligence - Top to Bottom | Ronald Petty | 27m |
Cloud Native AI means applying containers and orchestration to ML pipelines; the CN AI working group publishes a landscape and a paper, curates projects, and… | ||
| Lightning Talk: Beyond Polygons: The Future of 3D Graphics with Implicit Models | Pierre Wilmot | 20m |
Explains how photogrammetry rigs with about 70–75 cameras feed differentiable inverse rendering that trains volumetric models via gradient descent, and… | ||
| Composing AI Applications as a Graph with Haystack | Tuana Celik | 29m |
Haystack 2.0 models AI apps as a directed cyclic graph so pipelines can loop and become agentic; components are Python classes with run methods that pass typed… | ||
| Empowering AI Education with Rag: Building Domain-Specific LLM Agents for Enhanced... | Miley Fu | 38m |
Shows a working pipeline that turns course notes and textbooks into embeddings, loads them into a vector DB, and wires that RAG layer to self‑hosted LLMs… | ||
| Keynote: Common Corpus: Opening Data for Building Open Source LLMs | Anastasia Stasenko | 15m |
pleias launches the Common Corpus as a reproducible dataset of 1 trillion tokens first release drawn from public-domain cultural heritage (libraries, national… | ||
| ML Data Version Control and Reproducibility at Scale | Einat Orr | 38m |
Demonstrates a Git-like system (lakeFS) that version-controls S3/Minio data with zero-copy pointers so petabytes don’t get duplicated, supports branches that… | ||
| Protecting Open Innovation in Future Regulation | Ben Brooks | 42m |
Open models power huge grassroots R&D — Stable Diffusion family records ~150 million downloads and 300,000 Discord contributors and open fine‑tuning now… | ||
| Running and Fine-tuning Open Source LLMs on Google Kubernetes Engine | Nael Fridhi | 32m |
Deploys a Llama 2 70B model on Google Kubernetes Engine using Nvidia L4 GPUs, mounting model and checkpoints to Cloud Storage with GCSFuse, and serves via TGI… | ||
| Advanced RAG Techniques with LlamaIndex | Pierre-Loic Doulcet | 35m |
LlamaIndex presents concrete fixes for RAG pipelines: an open source LlamaIndex Python framework with recent releases like composable memory for agents, a… | ||
| Hugging Face Accelerate: Making Device-Agnostic ML Training and Inference Easy... | Zachary Mueller | 23m |
An open-source library that turns a raw PyTorch loop into distributed training and inference by changing four lines of code: create an accelerator, call… | ||
| Keynote: Building with Mistral | Sophia Yang | 13m |
Small Paris team launches the open-weight Mistral 7B (Sep 2023), follow-up mixture-of-experts models (8x7B and 22B) and commercial MrSmall/MrLarge offerings… | ||
| Keynote: Fireside Chat with Soumith Chintala, Co-Creator of PyTorch & Merve Noyan | Soumith Chintala, Merve Noyan | 32m |
History of PyTorch is condensed into why torch.compile works most of the time and what engineers actually change to make it work — rewrite ops or make graphs… | ||
| nothing matches | ||