AI’s Bright Future: Insights from ODSC East 2025 Podcast Minisodes

ODSC East 2025 once again delivered a powerhouse of AI insights, featuring a unique podcast episode recorded live with short interviews from some of the brightest minds in AI today. Across these minisodes, speakers explored cutting-edge topics ranging from AI agents, small language models, and AI risk management, to synthetic data, causal AI, and even social media algorithms. Here’s a look at each mini-interview and the key takeaways from these experts.

You can listen to the full podcast on Spotify, Apple, and SoundCloud.

Rob Bailey (COO & Co-founder, Crew AI) — The Rise of AI Agents

Rob Bailey discussed how AI agents are moving beyond hype into real enterprise adoption. His keynote emphasized that many organizations are already actively working with AI agents to accelerate workflows. The audience’s engagement, filled with personal stories of agent implementation, underscored how AI agents are becoming a practical solution, not just a trend. His central message: AI agents are here, they’re growing fast, and businesses want to move quickly with them.

Ivan Lee (CEO & Founder, Datasaur.ai) — Small Language Models Take Center Stage

Ivan Lee made a strong case for the future being dominated by specialized, smaller language models. He highlighted how models like Google’s Gemma and Chinese open-source models are allowing powerful AI applications on limited hardware, even running offline on laptops. With open-source catching up quickly to proprietary models, developers have an unprecedented range of options depending on privacy, cost, and cloud preferences.

Tony Kipkemboi (Developer Advocate, Crew AI) — Hands-On with AI Agent Building

Tony led a highly attended workshop teaching foundational AI agent building skills. Participants built agents that automatically analyzed code repositories and generated documentation — addressing a common developer pain point. His session highlighted how accessible agent-building has become, empowering developers to create useful AI-driven tools without needing advanced expertise.

Samuel Colvin (CEO, Pydantic) — Type-Safe Agents and Observability

Samuel focused on the importance of type safety and observability in building robust AI agents. Using Pydantic AI and Logfire, he demonstrated how type-safe agents can be refactored quickly and how observability tools help track complex AI workflows. He also touched on the evolving but powerful field of AI evaluations (evals), crucial for refining agent performance.

Rajiv Shah (Machine Learning Engineer, Contextual AI) — Hill Climbing Evaluations

Rajiv introduced the concept of “hill climbing” evaluations, encouraging iterative, methodical improvements to generative AI systems. As AI development becomes increasingly non-deterministic, especially with prompt engineering and RAG (retrieval augmented generation), structured evaluation frameworks are essential. His practical advice emphasized continual experimentation tied to stakeholder needs.

Sinan Ozdemir (AI Expert & Author) — Beyond Benchmarks in AI Evaluations

Sina discussed why benchmarks should be the beginning, not the end, of AI evaluation. He stressed the need for internal evaluation policies that define what “better” means for each organization. For multi-agent systems, evaluations must also consider workflow efficiency and tool utilization, which go far beyond simple LLM response accuracy.

Alexandra Ebert (Chief Trust Officer, Mostly AI) — The Power of Synthetic and Open Data

Alexandra delivered a keynote on how open data — aided by privacy-preserving synthetic data — is key to achieving AI’s full societal potential. She announced the Mostly AI Prize, a $100K competition to advance open synthetic data innovation. By addressing data accessibility while protecting privacy, synthetic data enables broader AI adoption for both research and enterprise use.

KJ Patlolla (Product Manager, Google Cloud Agent Garden) — Building Production-Ready AI Agents

KJ outlined the challenges of bringing AI agents into production at scale. He introduced Google’s open-source Agent Development Toolkit (ADK) and emphasized the need for robust operations (agent ops) as organizations move from prototype to live deployment. His key message: scalable AI agent development requires mature ops practices, continual monitoring, and human oversight.

Cal Al-Dhubaib & Lauren Burke-McCarthy (Further) — Balancing AI Value and Risk

Lauren and Kal tackled AI risk management and sustainable AI adoption. Kal argued for risk management as an enabler of trust rather than a blocker of innovation, while Lauren highlighted how involving stakeholders early fosters alignment and trust. Together, they emphasized collaborative risk governance as essential for successful enterprise AI initiatives.

Andre Franca (CTO, causaLens) — Causal AI for Better Decisions

Andre made the case that true decision-making AI requires understanding causality, not just correlation. As companies increasingly adopt agents for business decision support, causal AI helps identify why things happen, leading to more effective interventions. He predicts a major role for “causal agents” as agentic systems mature.

Noah Giansiracusa (Professor, Bentley University) — Decoding Social Media Algorithms

Noah revealed that beneath the surface, most major social media platforms share a common formula based on engagement metrics. His talk aimed to empower users to better understand and manage their social media experiences by demystifying these algorithms, offering agency rather than algorithmic frustration.

Allen Downey (Principal Data Scientist, PyMC Labs) — Time Series and Bayesian Statistics

Allen conducted workshops on time series analysis and Bayesian statistics using PyMC. Geared toward beginners, his sessions provided participants with tools and resources to confidently start working with advanced statistical methods for forecasting and data modeling.

Conclusion

ODSC East 2025 showcased the incredible diversity of today’s AI landscape. From pragmatic agent-building to sophisticated evaluations and cutting-edge ethical data practices, these minisodes captured the pulse of AI innovation. Whether you’re a practitioner, executive, or simply an AI enthusiast, the insights shared offer a roadmap to navigating AI’s exciting and fast-evolving future.