The Top ODSC East 2025 Virtual Sessions Available On-Demand
ODSC East 2025 is done, and we’ll have to wait another year to gather together live in Boston. In the meantime, recordings of our virtual sessions are now available on demand! All of the amazing keynotes, talks, and deep dives into topics with training sessions and workshops are available for purchase. Check out a few of the highlights with their slides below.
LLMs That Think: Demystifying Reasoning Models
Sara Zanzottera, Senior Developer, BNP Paribas
Zanzottera’s talk unpacked how modern LLMs approach “reasoning,” breaking down common misconceptions around symbolic logic versus emergent capabilities. She provided a framework for thinking about different reasoning tasks (e.g., deduction, analogical thinking) and evaluated model performance using specific prompts. It’s an essential primer for anyone evaluating LLM capabilities beyond simple completion.
The 2025 Shift to Smaller Models: Why Specialized AI Will Win
Ivan Lee, CEO, Datasaur
Lee makes a compelling case that smaller, specialized models will dominate the next wave of AI innovation. He emphasizes performance per watt, latency constraints, and deployment realities over “bigger is better” approaches. Particularly relevant for edge applications and enterprises seeking privacy-preserving solutions.
Unlocking the Power of Data Storytelling: Turning Insights into Impact
Joshua Brindley, Manager: Corporate & Commercial Analytics, GEMS Education
Brindley’s session explored how visual storytelling, narrative framing, and design thinking can improve data adoption across organizations. He walks through a process for transforming dashboards and reports into decision-focused communication tools, especially useful for analytics teams working with non-technical stakeholders.
The A to Z of Building AI Agents
Apoorva Joshi, Senior AI Developer Advocate, MongoDB
This session is a comprehensive tour of AI agent architecture, from planning to prompt orchestration. Joshi dives into tool use, memory management, and LLM interfacing with MongoDB Atlas as a real-time knowledge store. The examples are production-ready and provide an actionable reference for developers and ML engineers alike.
Applied Data Mesh Workshop for Scalable Data Platforms
Jay Sen, Director, Data Engineering, PayPal
Sen led a highly practical walkthrough on implementing data mesh principles using modern tooling. He breaks down key roles, responsibilities, and the governance structures needed to scale decentralized data ownership. The presentation also includes a GitHub-linked demo — perfect for practitioners seeking a hands-on entry point.
LLM & RAG Evaluation Playbook for Production Apps
Paul Iusztin, Senior AI Engineer / Founder, Decoding ML
Iusztin’s talk provided a step-by-step evaluation framework for retrieval-augmented generation (RAG) pipelines, including metrics, benchmarks, and pitfalls in real-world deployments. It’s especially valuable for teams optimizing model outputs for consistency, factuality, and latency in knowledge-rich applications.
Rami Krispin, Senior Manager — Data Science and Engineering, Apple
Krispin’s workshop demonstrates how LLMs can be used as interfaces to query structured datasets, without exposing raw SQL. The session includes architectural best practices for embedding models into BI workflows and highlights the trade-offs between accuracy, interpretability, and real-time access.
Getting Started with Multi-Agent Applications
Valentina Alto, Technical Architect, AI & App, Microsoft
This beginner-friendly session introduces the core components of multi-agent systems, focusing on communication protocols, autonomy vs. orchestration, and interaction patterns. Alto also shows how to build an agent ecosystem using tools like LangChain, semantic memory, and Azure OpenAI.
Evals for Supercharging Your AI Agents
Aditya Palnitkar, Staff Software Engineer, Meta
Palnitkar outlines a framework for stress-testing AI agents using adversarial examples, simulation loops, and hybrid metrics. He discusses tools that Meta uses internally for red teaming and QA, giving attendees a peek behind the curtain of enterprise-grade model validation.
Building Real-World Agentic Applications with LlamaIndex
Laurie Voss, VP, Developer Relations, LlamaIndex
Voss illustrates how developers are building end-to-end agentic apps using LlamaIndex. Topics include routing, context windows, custom memory modules, and managing multi-step tasks. If you’re building copilots, chatbots, or autonomous research tools, this slide deck is a solid launchpad.
What’s Next: On-Demand Access & Future Events
ODSC East 2025 may be wrapped up, but many of the sessions are available on demand for registered attendees. Looking ahead, ODSC West 2025 in October and the highly anticipated Agentic AI Summit in July will offer even more technical deep dives into LLMs, agents, and enterprise AI workflows.
Visit ODSC.com to sign up for notifications, get early access to speaker lineups, and explore additional slides, code repos, and session recordings.
