AI By the Bay 2025
| title | author | length |
|---|---|---|
| The present and future of programming with AI (Keynote) | Peter Norvig | 26m |
Progress from 2023 AlphaCode to 2025 shows LLMs go from messy contest solutions to clean, efficient code (Claude reproduces AlphaCode’s backspace solution with… | ||
| Future of AgenticAI: Architecting the Global Control Plane (Keynote) | Tyler Jewell | 30m |
Company-grade platform reorients agentic AI around a distributed control plane not single agents, shipping an SDK, cluster runtime and event‑sourced… | ||
| From Data 2.0 to Data 3.0: Making Data Autonomous for AI(Keynote) | Zhamak Dehghani | 26m |
Claims a new runtime abstraction called Data 3.0 that packages data plus computation into autonomous data products which run, self-orchestrate, enforce… | ||
| LLMs & Data Warehouses: Reliable AI Agents for Business (Keynote) | Ryan Boyd | 40m |
LLMs often hallucinate and fail at data aggregation so the talk demonstrates concrete fixes: use semantic models or commented views, feed a DuckDB/MotherDuck… | ||
| Durable Agents: long running AI workflows in a flakey world | Samuel Colvin | 24m |
Shows a concrete system: Pydantic AI plus a Temporal workflow wraps every LLM or tool call as an activity so runs survive node crashes and network flakiness… | ||
| Agents + Data Streaming for Real-Time Enterprise Intelligence | Peter Corless | 25m |
A reference architecture proposes MCP-driven, A2A agent systems that interrogate real-time streams across five industries — financial services, cybersecurity… | ||
| Challenges of Ambient Agents: Building Governable AI Systems at Scale | Fred Patton | 27m |
Ambient agents must be fast, cheap, and governable simultaneously, so the concrete solution is to convert raw event streams into actionable state using stream… | ||
| Lakeside AI: Bringing Models to Data with Open Lakehouses & Agents) | Mainak Ghosh, Martin Traverso, Jitender Aswani | 29m |
Claim: move models to data by ingesting streaming and batch sources once into Apache Iceberg and serving many AI uses with Trino-powered SQL and agent hooks… | ||
| Taming ML’s Wild West: Practical Model Signing with Sigstore on Kaggle | Mihai Maruseac | 26m |
Practical steps for attesting model integrity: use the Python library model-signing (pip install model-signing) with Sigstore to create short‑lived OIDC-backed… | ||
| Automatic Agent Optimization and the Rise of Self-Improving Agents | Vincent Koc | 33m |
Introduces Opik Agent Optimizer SDK and two optimizer classes—an evolutionary/genetic optimizer (with population, mutation, hall of fame and Pareto scoring)… | ||
| From Experiment to Enterprise: How Block Operationalized MCP at Scale | Angie Jones | 30m |
Block deploys MCP organization-wide in about two months, building an open-source MCP client called Goose that connects to Snowflake, Jira, PagerDuty and other… | ||
| Agents Need More than Vector Search | Prashanth Rao | 31m |
LanceDB presents a multimodal lakehouse that combines vector and full‑text indices, persisted to disk or object stores, and treats images/audio/video as… | ||
| Context Engineering: Connecting the Dots with Graphs | Stephen Chin | 30m |
Neo4j Aura hosts a 135-node, 418-relationship knowledge graph built from a BOM and a VEX file using an open-source knowledge graph builder, then used two… | ||
| Graphs as the Backbone of Scalable Agentic Apps | Laura Gutiérrez Funderburk | 30m |
Compares using haststack pipelines as deployable microservices for indexing and retrieval with Langraph as an orchestrator that only holds state and chooses… | ||
| Eclipse LMOS&AGC: Agentic Computing for Real-World Enterprise Systems | Arun Joseph | 33m |
Built for a Deutsche Telekom deployment across ~10 countries and ~200–300 APIs, Eclipse LMOS provides a Kotlin framework (LMOS Arc) plus a declarative ADL… | ||
| Safety in Foundational Models | Vinay Rao | 34m |
Claims that frontline providers already run bioweapon detection classifiers and that cloud code is being used to orchestrate cyber attacks at ~80% autonomy… | ||
| MCP and the Dawn of Agentic Intelligence | Gabriela de Queiroz, Shub Argha | 12m |
MCP claims to turn LLMs into context-aware agents by standardizing client–server capability negotiation for tool chaining, resource access, sampling and… | ||
| Verifiably Safe Agentic Compute | Erik Meijer | 33m |
Claims that LLMs are just branch predictors and that useful agentic systems need a new compute stack not bigger models. Proposes a programming language that… | ||
| The advent of the open data lake | Julien Le Dem | 33m |
Ecosystem shifts from Hadoop-style bundled stacks to composable layers that let teams own storage in blob stores and swap engines: columnar files like Parquet… | ||
| The World's Largest Open Source Entity Graph | Jose Plehn | 36m |
An open entity graph is released under CC0 license covering about 300 million companies, 500 million places and roughly 1.25 billion people with edges linking… | ||
| AI For Good: Fighting Health Insurance with AI | Holden Karau | 31m |
A nonprofit pipeline turns published state independent medical review decisions into training pairs, fine-tunes Gemma/Mistral-derived models, and serves them… | ||
| Designing Data Infrastructure in the Age of Generative AI | Lisa N. Cao | 32m |
Delta kernel is a Rust library that embeds the Delta protocol into engines so multiple runtimes can coordinate commits via Unity Catalog instead of per-format… | ||
| BAML - A new programming language for building Agents | Vaibhav Gupta | 36m |
BAML is a domain-specific language that embeds inside any host language and turns LLM requests into type-safe LM calls with generated streaming and… | ||
| Identity Best Practices to Secure & Scale Agentic AI / MCP Deployments | Kevin Gao | 23m |
More than 5,000 MCP servers appear in registries yet over 2,000 expose no security and only about 13% ever deploy, so the core takeaway is concrete: implement… | ||
| Scaling Multimodal Data and Reinforcement Learning with Ray | Jaikumar Ganesh | 28m |
Claims that a Kubernetes + Ray + PyTorch + vLLM stack handles heterogeneous AI workloads, with Ray providing finer-grain scheduling, failure recovery, and… | ||
| Uber's Multi-Agent Platform | Jamieson Leibovitch | 31m |
Uber claims single agents aren’t enough and ships a production stack: AgentFx SDK (internal agent runtime with a LangChain-linked agent protocol fork of… | ||
| Trusted Agentic AI:Architecting Systems with Reliable Data at Scale | Steve Anderson | 23m |
Claims an AI radiology pipeline finds lesions in 10,000 abdominal images per week and raises cancer survival by earlier detection; an agentic analytics product… | ||
| Introducing Weave-CLI: The easiest way to manage your vector DB | Max Maximilien | 33m |
A fast Go CLI that manages Weaviate, Milvus and Supabase vector stores, handling collection creation, chunking, embeddings, listing, JSON search results and… | ||
| Reliable Agentic AI, the Open Way | Mark Collier | 14m |
AI shifts from chat to systems and the talk lays out three concrete pillars: training, inference, agents, with specific claims and examples. Infrastructure… | ||
| From Growth Hacking to Vibe Coding and Back Again | Joel S. Horwitz | 33m |
Charts a concrete shift from hand-built ML experiments to Vibe Coding and agentic growth: builds Center Media with AMP Code, connects to Clay, PostHog… | ||
| Beyond GPUs: Why AI Agents Need a Need New Hardware | Vasanth Mohan | 30m |
Argues that GPUs hit an efficiency wall for agentic AI and shows a different chip architecture: a reconfigurable data‑flow unit with a three‑tier memory on the… | ||
| Is Generative BI failing? | Boris Renski | 14m |
Argues GenBI stalls because of an unrealistic “ask anything, get magic” promise and users expecting one-shot answers; pilots mostly die. Builds a RAG +… | ||
| Agentic Coding with Discipline & Skill: Practical Techniques and Tips | YK Sugi | 27m |
Builds and debugs production workflows using agent tools like Claude Code and an earlier local-file plugin called Kaguya, claiming hundreds of millions of… | ||
| Building Tools with Tools: Modern AI-Driven Development | Edir Garcia Lazo | 30m |
Agents sit at the front of the SDLC and the workflow replaces slow manual implementation with a loop of idea→GitHub issue→agent→review→prod, using a text… | ||
| RAG + Apache Iceberg™: Structured Retrieval Made Easy | Josh Reini, Vino Duraisamy | 33m |
DonutBot maps free‑form donut orders to real menu items by combining semantic vector search over product descriptions with exact filters pulled from Apache… | ||
| Build Better AI with First-Party Data and Fine-Tuning | Paul Yang | 33m |
Claim that small fine-tuned models on first‑party data beat giant LLMs on domain tasks with concrete numbers: Fashion MNIST examples show GPT‑4 mini ~81% vs… | ||
| Event-Driven & Multi-Agent Architectures for Gen AI Workflows | Mary Grygleski | 32m |
Argues that combining event sourcing + CQRS with multi‑agent orchestration turns one‑shot LLM calls into reproducible, scalable workflows; demonstrates using… | ||
| From Big Data to DuckDB: Ryan Boyd (MotherDuck) on Hyper-Tenancy, Duck Lake, and Making Data Fun. | Ryan Boyd | 38m |
Argues that replacing massive distributed engines with many small DuckDB instances yields real wins: MotherDuck’s hyper-tenancy gives each customer or user a… | ||
| Akka - Enterprise Agentic AI with Tyler Jewel | Tyler Jewel | 28m |
Akka has pivoted into a platform for running stateful agentic systems at enterprise scale and claims 45 customers in production plus a recent competitive win… | ||
| Francis Bouvier | Francis Bouvier | 15m |
Identifies three risks—credential exposure, shared remote browser sandboxes, and prompt injection—and proposes running a local lightweight browser per task… | ||
| Full-Stack AI Engineering from AI-Native Coding to Knowledge, Models, Infrastructure & OSS | Alexy Khrabrov | 14m |
Conference lays out concrete claims and failures: modern dev tools like copilots and repet can produce clickable code that breaks test workflows when dynamic… | ||
| How to Write a Strong Conference Proposal? | Alexy Khrabrov, Lisa N. Cao, Baruch Sadogursky | 46m |
Practical checklist: start with a one‑sentence thesis like X because Y results in Z, state the specific problem and the concrete takeaway, and name exact tools… | ||
| Context is All You Need: The Art and Science of Information Design | Lena Hall | 30m |
Claim: failures come from bad information design not weak models, so reliability requires persistent, versioned context and reviewable plans. Empirical… | ||
| The AI Operating System: Stateful Agents with Letta | Cameron Pfiffer | 32m |
Letta claims to turn stateless LLM interactions into stateful agents that persist, edit memories and self-improve; it implements a memory-first agent design… | ||
| The Missing Protocol: How MCP Bridges LLMs and Data Streams | Viktor Gamov | 35m |
Claims that MCP turns LLMs into first‑class participants in streaming architectures by exposing Kafka and other systems as callable tools. The demo wires an… | ||
| Meet Jules, Your Autonomous Coding Partner! | Oleg Aravin | 31m |
Demo shows Jules running on Google cloud with Gemini 2.5, booting a VM, cloning a repo and producing ready pull requests that build and preview a site. It uses… | ||
| Enterprise-ready document processing for AI | Kerim Satirli, Tu Nguyen | 39m |
Built an end-to-end pipeline that converts PDFs, images, audio and video into chunked embeddings, stores them in a vector DB and serves relevant passages to… | ||
| Building memory for AI Agents | Vasilije Markovic | 21m |
Cognee is a Python SDK that connects LLMs to graph and vector stores and claims to solve memory consistency at scale by combining Redis short‑term cache… | ||
| From Bots to Brains: The Evolution of Web Scraping | Lorenzo Padoan | 10m |
A history-to-future survey that claims web scraping is shifting from brittle rule-based crawlers (BeautifulSoup, regex, Selenium/Playwright, proxy evasion) to… | ||
| Can We Trust AI-Generated Code? | Baruch Sadogursky, Leonid Igolnik | 32m |
Claims that LLMs produce correct code only about half the time (slide cites ~50% correctness and ~40% of AI code vulnerable) and that self-verification creates… | ||
| Spec Driven Development: AI Assisted Coding With Doc at the Core | Bruno Borges | 37m |
Claims that structured, machine-readable specs make AI outputs deterministic and reviewable and shows a concrete toolchain: GitHub Spec Kit plus Visual Studio… | ||
| Open Standards and Lakehouse AI Development | Alex Merced | 31m |
Argues that combining Apache Iceberg table metadata, the Apache Iceberg REST Catalog Spec and columnar transport via Arrow Flight yields a lakehouse that is… | ||
| Context engineering in ever changing environment - made simple | Linghua Jin | 15m |
An intern’s cron job that reprocesses every file hourly wastes compute when only 1% of content changes and creates ghost records when filenames or offsets… | ||
| Live Demo – Develop and Deploy a Trustworthy Multi-Agent System | Tyler Jewell | 30m |
Live demo builds a Java/Maven Akka-based agentic platform that compiles locally into the same binaries used in production, runs a three-node cluster with… | ||
| nothing matches | ||