ODSC's Ai X Podcast

ODSC

With Ai X Podcast, Open Data Science Conference (ODSC) brings its vast experience in building community and its knowledge of the data science and AI fields to the podcast platform. The interests and challenges of the data science community are wide ranging. To reflect this Ai X Podcast will offer a similarly wide range of content, from one-on-one interviews with leading experts, to career talks, to educational interviews, to profiles of AI Startup Founders. Join us every two weeks to discover what’s going on in the data science community. Find more ODSC lightning interviews, webinars, live trainings, certifications, bootcamps here - https://aiplus.training/ Don't miss out on this exciting opportunity to expand your knowledge and stay ahead of the curve.

  1. Designing MCP for Real-World Agents with Jeremiah Lowin

    20 THG 5

    Designing MCP for Real-World Agents with Jeremiah Lowin

    In this episode, Sheamus McGovern speaks with Jeremiah Lowin, Founder and CEO of Prefect and creator of FastMCP, for a deep dive into how Model Context Protocol (MCP), is evolving from simple tool exposure into real infrastructure for enterprise AI agents. Jeremiah explains why many MCP servers are poorly designed, why internal enterprise adoption has become MCP’s strongest product-market fit, and how emerging patterns like dynamic tool search, code mode, and MCP Apps are changing what is possible for agentic systems in production. This episode was recorded before ODSC AI East 2026, which has passed. Key Topics Covered Why are many MCP servers failing Why the strongest product-market fit for MCP today is inside the enterprise Common internal enterprise MCP use cases How MCP is helping organizations move closer to the long-promised idea of the self-serve data platform, while also creating the need for guardrails Design principles for MCP servers Why conventional REST APIs often make poor MCP servers How MCP client improvements are changing best practices, especially with dynamic tool search and smarter clients that no longer need to load all tools into context at once. What code mode is, how it works, and why it matters Jeremiah’s framing of MCP as a substrate connecting agents to companies, What MCP Apps are and why they represent an important shift beyond tool calling How MCP Apps and generative UI could change the way humans and agents work Why organizations need executable, observable workflows A practical view of enterprise agent autonomy, The shift in MCP architecture from many tiny constrained servers toward fewer Where OpenClaw fits into the broader MCP discussion Memorable Outtakes “ if you have an MCP server with 1,000 tools on it, Claude can handle it no problem” “MCP has allowed is a clear separation of the business logic and the agentic loop.” “People treat the LLMs like they’re oracles as opposed to interns.” References & Resources Jeremiah Lowin LinkedIn: https://www.linkedin.com/in/jlowin/ FastMCP GitHub: https://github.com/prefecthq/fastmcp FastMCP documentation: https://gofastmcp.com Prefect: https://www.prefect.io MCP Apps : https://modelcontextprotocol.io/extensions/apps/overview ODSC West: https://odsc.ai/west Sponsored by This episode was sponsored by: ODSC AI West 2026 – The Leading AI Builders Conference Join us in San Francisco from October 27th–29th for expert-led sessions on Agentic AI, AI Engineering, Data Science, Machine Learning, LLMOps, and AI-driven automation. Learn more: https://odsc.ai/west ODSC AI Engineering Accelerator – From Using AI to Building It Seven weeks of live, cohort-based training in LLMs, AI assistants, agents, and AI engineering. Ship a working AI agent or assistant by the end. Summer and Fall cohorts available. Group pricing for teams of 3+. Learn more: https://odsc.ai/west/accelerator

    50 phút
  2. How AI is Transforming the Fan Experience in Major League Baseball with Neil Weiss

    12 THG 5

    How AI is Transforming the Fan Experience in Major League Baseball with Neil Weiss

    In this episode of the ODSC Ai X Podcast, host Cal Al-Dhubaib speaks with Neil Weiss, Executive Vice President, CIO, Government Affairs & Economic Development for the Cleveland Guardians. Neil discusses how the Guardians use technology, data, AI, and operational innovation to compete against larger-market MLB teams, improve the fan experience, support internal teams, and help the organization make smarter decisions across both business operations and player development. The conversation explores everything from AI-powered customer feedback analysis and agentic self-service tools to biomechanics, high-speed video, organizational learning, and the importance of keeping people at the center of technology adoption. Key Topics Covered: Neil’s unconventional role across IT, government affairs, economic development, and ballpark operations at the Cleveland Guardians How the Guardians use technology as a competitive advantage in a smaller-market MLB environment Why the organization’s “win more with less” mindset applies to both baseball operations and business operations How data supports fan safety, incident management, work order management, and operational decision-making inside the ballpark The use of incident data to identify dehydration risks, optimize cooling stations, and improve the game-day experience How the Guardians’ consumer research and insights team uses AI to analyze large volumes of fan survey feedback more efficiently Why parking can be a hidden root cause of broader fan dissatisfaction, and how connected data can uncover those patterns How AI agents and self-service tools are changing ticket renewals, customer service, and fan communication Why automation should free up human teams to spend more time on high-value, relationship-driven interactions Neil’s approach to AI governance, including policy, security, protecting IP and PII, and encouraging experimentation without creating unnecessary risk The importance of starting with business outcomes rather than starting with tools or technology Why time savings from AI only matter if organizations intentionally reapply that time toward meaningful goals How the Guardians are training employees on generative AI through recurring internal and external education sessions The organization’s plans to take bigger AI-driven swings around critical business processes, knowledge retrieval, finance, people operations, and ballpark operations How the Guardians are exploring a more product-driven approach to fan-facing innovation, moving beyond the traditional season-by-season planning cycle The role of computer vision, high-speed video, biomechanics, and data modeling in player development How complex data insights are translated into practical coaching guidance for players Why Neil believes technology and data should enable human beings rather than replace them How Neil stays current on AI and technology through podcasts, industry publications, peer networks, and continuous learning Memorable Outtakes: “Technology is a huge leverage point for us, and it has been for a long time, and it continues to be, and it always will be.” — Neil Weiss “You gotta talk about process and people and outcomes. And if you can apply technology, great, but don’t do it just for technology’s sake because it seems cool.” — Neil Weiss “There aren’t computers doing the work. There’s computers and data enabling human beings to do the work.” — Neil Weiss References & Resources: Neil Weiss on LinkedIn: https://www.linkedin.com/in/weissneil/ Cleveland Guardians: https://www.mlb.com/guardians The AI Daily Brief: https://aidailybrief.ai/ Sponsored by: This episode was sponsored by: ODSC AI West 2026 – The Leading AI Builders Conference Join us in San Francisco from October 27th–29th for expert-led sessions on Agentic AI, AI Engineering, Data Science, Machine Learning, LLMOps, and AI-driven automation. Learn more: https://odsc.ai/west

    49 phút
  3. Engineering the Agentic Harness with Rajiv Shah

    6 THG 5

    Engineering the Agentic Harness with Rajiv Shah

    In this episode, we speak with Rajiv Shah, an Agentic AI Engineer at OpenHands, about the agentic harness behind modern AI systems, which entails the retrieval, memory, tools, orchestration, and execution environment that determine whether an AI agent can work reliably in practice. Rajiv shares practical insights on context engineering, agentic RAG, memory design, orchestration patterns, and the tradeoffs teams need to understand when building reliable AI agents. Note: This episode was recorded and edited before ODSC AI East 2026. Key Topics Covered: What the “harness” means in agentic AI, and why it includes retrieval, memory, tools, orchestration, and execution environment Why retrieval remains the foundation of reliable AI systems How RAG has evolved from dense retrieval to hybrid approaches that combine lexical and embedding-based search What context engineering means in practice for AI engineers and practitioners How agentic RAG allows models to reformulate queries, loop through retrieval, and improve search results How to think about short-term context, long-term memory, and persistent artifacts Planner-executor patterns, supervisor routing, and how teams structure longer-running agent workflows Why multi-agent systems often add coordination costs, latency, and complexity How agent-to-agent handoffs work in practice, often through APIs and JSON rather than formal protocols Why overengineering is one of the biggest misconceptions in agentic AI right now Memorable Outtakes: - “All of those components that help to take our coding model to actually be able to solve coding tasks become that part of the harness.” - “You wanna stay with a single agent as long as possible.” References & Resources: - Rajiv Shah’s LinkedIn: https://www.linkedin.com/in/rajistics/ - OpenHands Official Site: https://openhands.dev - OpenHands GitHub: https://github.com/OpenHands/OpenHands - Contextual AI Official Site: https://contextual.ai/ This episode was sponsored by: ODSC AI West 2026 – The Leading AI Builders Conference Join us in San Francisco from October 27th–29th for expert-led sessions on Agentic AI, AI Engineering, Data Science, Machine Learning, LLMOps, and AI-driven automation. Learn more: https://odsc.ai/west

    38 phút
  4. LLMs Aren't Enough: Teaching Agent Swarms to Work Together with Nouha Dziri

    29 THG 4

    LLMs Aren't Enough: Teaching Agent Swarms to Work Together with Nouha Dziri

    In this episode, Sheamus McGovern sits down with Nouha Dziri, Research Scientist at Cohere and an award-winning AI researcher whose work has been recognized at top conferences like NeurIPS and NAACL. Nouha’s research focuses on understanding the limits of large language model reasoning, how reinforcement learning can push beyond those limits, and what comes next with multi-agent systems. The conversation explores why today’s models can solve complex problems yet fail on simple ones, revealing the gap between generation and true understanding. From there, the discussion shifts to a bigger question: what comes next if LLMs alone aren’t enough. Nouha explains why the future of AI lies in systems of agents that must learn to coordinate through training rather than rely on engineered workflows. Key Topics Covered: Why LLMs succeed on complex tasks but fail on simple or unfamiliar ones Hallucinations and the concept of “jagged intelligence” The difference between pattern matching and true reasoning in LLMs The “Generative AI Paradox”: models generate better than they understand or verify Reinforcement Learning vs Supervised Fine-Tuning (SFT) The importance of reward design, infrastructure, and strong base models in RL The RL Grokking phenomenon: sudden performance jumps after exploration Why single-agent systems are hitting limits on complex tasks Multi-agent systems as the next phase of AI development Limitations of current agent frameworks (LangGraph, CrewAI, orchestration patterns) Training agents to coordinate: communication, negotiation, and division of labor AGI as a system of interacting agents rather than a single model Emergent behavior in multi-agent systems (collusion, coordination failures) Safety risks in multi-agent systems: The Artificial Hivemind: collapse of diversity in model outputs Memorable Outtakes: “Models hallucinate because they don’t know when they’re wrong.” “Intelligence should grow more like a city than a solitary brain.” “Being trained to coordinate and just reading about coordination are very different things.” References & Resources: Nouha Dziri – LinkedIn: https://www.linkedin.com/in/nouha-dziri-3587427b/ Nouha Dziri – Publications: https://nouhadziri.github.io/publications/ Key Papers & Work Mentioned: Faith and Fate (NeurIPS 2023):https://arxiv.org/abs/2305.18654 Generative AI Paradox (ICLR 2024) https://arxiv.org/abs/2311.00059 RL Grokking Recipe (COLM 2026): https://tinyurl.com/ntarc3kw WildTeaming & WildGuard (NeurIPS 2024): https://arxiv.org/abs/2406.18495 Artificial Hivemind (NeurIPS 2025 Best Paper): https://arxiv.org/abs/2510.22954 Additional Concepts & Tools Referenced: MoltBot (agent social network experiment): https://moltbot.com OpenClaw (multi-agent system experiment) https://openclaw.ai LangGraph: https://www.langchain.com/langgraph CrewAI: https://www.crewai.comu Sponsored by: This episode was sponsored by: 🔥 ODSC AI East 2026 – The Leading AI Training Conference Join us in Boston from April 28th–30th for expert-led sessions on Agentic AI, AI Engineering, Data Science, Machine Learning, LLMOps, and AI-driven automation. Learn more: https://odsc.ai/east

    51 phút
  5. How AI Governance is Changing the Data Science Role with Shoshana Rosenberg

    21 THG 4

    How AI Governance is Changing the Data Science Role with Shoshana Rosenberg

    In this episode of the ODSC Ai X Podcast, host Alex Landa sits down with Shoshana Rosenberg—author of Practical AI Governance, Managing Director of Logical AI Governance, and Co-Founder of Women in AI Governance—to unpack what AI governance actually means in 2026. Drawing on her background spanning law, engineering, and enterprise strategy, Shoshana reframes AI governance as a dynamic business intelligence function rather than a static compliance exercise. The conversation explores how the role of data scientists is evolving, why evaluation and monitoring are becoming core disciplines, and how organizations must adapt to a constantly shifting AI landscape. Key Topics Covered: The definition of AI governance as a business intelligence and strategy function, not just compliance Why static governance frameworks fail in a rapidly evolving AI ecosystem The expanding role of data scientists beyond model building into evaluation, monitoring, and translation The growing importance of qualitative evaluation, red teaming, and adversarial testing How data scientists can bridge technical outputs with business strategy and organizational priorities Responsibility boundaries: bias, fairness, and safety as shared organizational concerns The intersection of regulatory literacy and technical safeguards Domain expertise vs. generalist skills in modern AI roles Evaluation as an ongoing discipline rather than a one-time checkpoint Risks of mistaking fluency (LLMs) for true understanding The rise of AI systems practitioners focused on assembling, monitoring, and governing AI systems Why adaptability—not fixed rules—is the foundation of effective AI governance Memorable Outtakes: “AI governance is not a compartmentalized compliance program—it’s a targeted business intelligence program.” “If you take aim at a fixed state, you will miss the path that you’re actually on.” “It’s no longer just about delivering a model—you don’t deliver and walk away anymore.” “Evaluation is not a gate. It’s an ongoing discipline.” References & Resources: LinkedIn: https://www.linkedin.com/in/shoshanarosenberg/ Book, “Practical AI Governance”: https://www.practicalaigovernance.com/ Safe Porter: https://www.safeportersecure.com/ AI DEI Privacy: https://www.aideiprivacy.com/ Logical AI Governance: https://www.logicalaigovernance.com/ Women in AI Governance (WiAIG): https://www.wiaig.com/ Sponsored by: 🔥 ODSC AI East 2026 – The Leading AI Training Conference Join us in Boston from April 28th–30th for expert-led sessions on Agentic AI, AI Engineering, Data Science, Machine Learning, LLMOps, and AI-driven automation. Learn more: https://odsc.ai/east

    38 phút
  6. From Annotation to AI Tutor: Who’s Training AI Now? With Jason Corso

    15 THG 4

    From Annotation to AI Tutor: Who’s Training AI Now? With Jason Corso

    In this episode, Sheamus McGovern speaks with Jason Corso, Professor of Electrical Engineering and Computer Science at the University of Michigan and Co-Founder & Chief Scientist at Voxel51. Jason is a leading expert in computer vision and machine learning, with deep experience in data-centric AI systems. The conversation explores the critical role of annotation in building modern AI models, how data—not just algorithms—drives performance, and how annotation workflows are evolving with foundation models. Jason also shares practical frameworks for dataset curation, evaluation, and human-in-the-loop systems, along with insights into the future of AI work, including the rise of AI tutors and the growing importance of data expertise. Speaker: Jason Corso University Profile: https://web.eecs.umich.edu/~jjcorso/ Voxel51: https://voxel51.com FiftyOne (Open Source Tool): https://github.com/voxel51/fiftyone LinkedIn: https://www.linkedin.com/in/jason-corso/ Key Topics Covered: What annotation is and why it is foundational to modern AI systems How labeled data and supervised learning powered the AI boom The role of self-supervision and how LLMs changed labeling dynamics Model-assisted annotation and shifting human roles toward QA and edge cases The importance of dataset curation and understanding your data before training Why data quality and selection matter more than model choice in many applications The role of scenario analysis and identifying failure modes in production systems Why gold-standard datasets must evolve over time rather than remain static Techniques like active learning, weak supervision, and synthetic data How foundation models can be used for auto-labeling and bootstrapping datasets The challenge of long-tail data and rare edge cases in real-world systems The importance of iterative data loops: train → evaluate → analyze → improve The emergence of new roles like AI tutors and expert annotators Why AI is more likely to augment than replace white-collar work Memorable Outtakes: “Annotation is the fundamental aspect of the success we’ve seen in modern AI.” “The performance you get is ultimately a function of the data you use—not just the model you choose.” “AI is going to be a boon for white-collar work—not a replacement for it.” References & Resources: Jason Corso (University of Michigan): https://web.eecs.umich.edu/~jjcorso/ Voxel51: https://voxel51.com FiftyOne (Open Source Dataset Curation Tool): https://github.com/voxel51/fiftyone ImageNet Dataset (large-scale labeled dataset that enabled deep learning breakthroughs) https://www.image-net.org MS COCO Dataset (segmentation and object detection benchmark) https://cocodataset.org LVIS Dataset (large-scale dataset with fine-grained categories) https://www.lvisdataset.org Computer Vision Annotation Tool (CVAT) https://www.cvat.ai This episode was sponsored by: ODSC AI East 2026 : https://odsc.ai/east– The Leading AI Builders Conference Join us in Boston from April 28th–30th for expert-led sessions on Agentic AI, AI Engineering, Data Science, Machine Learning, LLMOps, and AI-driven automation. Learn more: https://odsc.ai/east

    49 phút
  7. Using AI to Make Healthcare More Human with David Sylvan

    7 THG 4

    Using AI to Make Healthcare More Human with David Sylvan

    In this episode of the ODSC Ai X podcast, Cal Al-Dhubaib speaks with David Sylvan, Chief Strategy, Innovation & Marketing Officer at University Hospitals, about how AI is moving from experimentation to real-world impact in healthcare. David shares how his team approaches AI investment, governance, and adoption in a highly regulated environment, and why success depends on solving real problems, involving end users early, and focusing on both patient experience and caregiver workflows. The conversation also explores how University Hospitals is using generative AI to improve efficiency, reduce burnout, and create more human-centered care. Key Topics Covered: David Sylvan’s role at University Hospitals and how strategy, innovation, marketing, and commercialization work together Why healthcare organizations should start with real business and care delivery problems, not AI hype How University Hospitals prioritizes AI investments across patient experience, caregiver needs, and operational efficiency Lessons from AI initiatives in radiology, innovation, and sports health The role of UH Ventures, the Veale Healthcare Transformation Institute, and the Haslam Sports Innovation Center How generative AI is already being used to support patient outreach and medication-related workflows Why trust, governance, and stakeholder alignment are essential for AI adoption in healthcare How AI can reduce burnout and help clinicians work at the top of their license Why David believes AI can ultimately bring more humanity back into healthcare delivery What the next 12 months may look like as healthcare systems operationalize AI at scale Memorable Outtakes: David Sylvan: “We start by inventorying problems. We don’t try to start with the conjuring of a solution.” David Sylvan: “The appropriate adoption of AI can help bring humanity back into the delivery of healthcare.” Cal Al-Dhubaib: “The option of ‘I don’t like it, put it back,’ is just not on the table.” References & Resources: David Sylvan, University Hospitals: https://www.uhhospitals.org/about-uh/leadership/uh-system-leadership/david-sylvan David’s LinkedIn: https://www.linkedin.com/in/davidsylvan/ University Hospitals: https://www.uhhospitals.org/ The Veale Healthcare Transformation Institute: https://www.uhhospitals.org/veale-institute UH Haslam Sports Innovation Center: https://uhhaslamsportsinnovation.org/ Hippocratic AI: https://hippocraticai.com/ Sponsored by: This episode was sponsored by: ODSC AI East 2026 – The Leading AI Training Conference Join us in Boston from April 28th–30th for expert-led sessions on Agentic AI, AI Engineering, Data Science, Machine Learning, LLMOps, and AI-driven automation. Learn more: https://odsc.ai/east

    46 phút
  8. From AI Hype to Enterprise Execution: How to Actually Scale AI in 2026 with Linda Yao

    27 THG 3

    From AI Hype to Enterprise Execution: How to Actually Scale AI in 2026 with Linda Yao

    In this episode of the ODSC Ai X Podcast, host Cal Al-Dhubaib sits down with Linda Yao, Vice President and General Manager of Hybrid Cloud and AI Solutions at Lenovo Services Group, an $8B global business operating across 180+ countries. Linda shares how enterprises can move beyond AI experimentation into real-world execution by focusing on hybrid AI strategies, governance, and organizational readiness. Drawing from Lenovo’s own internal AI journey, she explains why success depends not just on models, but on systems, people, and process—and what it truly takes to scale AI securely and effectively in today’s enterprise landscape. Key Topics Covered: What “hybrid AI” means and why bringing AI to your data is critical for enterprise success The gap between AI excitement and real-world execution in organizations Why most enterprises are stuck in pilot mode and how to move toward production The importance of data quality, governance, and security as foundational AI requirements Differences in perspective between CEOs, CFOs, and CIOs when adopting AI Real-world enterprise AI use cases, including supply chain optimization and contact centers Human + AI collaboration: augmenting employees rather than replacing them Building trust in AI through responsible AI principles, transparency, and governance The role of organizational readiness across people, process, data, and infrastructure How AI adoption is driving improvements in data management and operational discipline Why orchestration—not just models—is the true source of competitive advantage in AI The rise of agentic and physical AI systems and what’s coming next Career lessons from three waves of enterprise AI and how to take advantage of disruption Memorable Outtakes: “One of the biggest lessons we’ve learned is that AI success is never just about the model or the technology—it’s about all the systems and people around it.” “There’s a lot of excitement around AI, but not always a clear path on how to scale it—especially in a secure and sustainable way.” “Even with the best conversational AI, you can’t talk your way to AI success.” References & Resources: - Linda Yao’s LinkedIn: https://www.linkedin.com/in/lindayao/ - Lenovo news + blog: https://news.lenovo.com/ - Lenovo AI services: https://www.lenovo.com/hk/en/services/ai-services Sponsored by: This episode was sponsored by: ODSC AI East 2026 – The Leading AI Training Conference Join us in Boston from April 28th–30th for expert-led sessions on Agentic AI, AI Engineering, Data Science, Machine Learning, LLMOps, and AI-driven automation. Learn more: https://odsc.ai/east Use the code podcast for an additional 10% off.

    45 phút
Xem tất cả (114)

Xếp Hạng & Nhận Xét

5
/5
5 Xếp hạng

Giới Thiệu

With Ai X Podcast, Open Data Science Conference (ODSC) brings its vast experience in building community and its knowledge of the data science and AI fields to the podcast platform. The interests and challenges of the data science community are wide ranging. To reflect this Ai X Podcast will offer a similarly wide range of content, from one-on-one interviews with leading experts, to career talks, to educational interviews, to profiles of AI Startup Founders. Join us every two weeks to discover what’s going on in the data science community. Find more ODSC lightning interviews, webinars, live trainings, certifications, bootcamps here - https://aiplus.training/ Don't miss out on this exciting opportunity to expand your knowledge and stay ahead of the curve.

Thông Tin

  • Nhà sáng tạo
    ODSC
  • Năm hoạt động
    2023 - 2026
  • Tập
    114
  • Trang web chương trình
    ODSC's Ai X Podcast

Có Thể Bạn Cũng Thích