In this episode, we sit down with Brian Lucena, Principal at Numeristical, and an experienced educator, consultant, and open-source contributor. Brian has advised companies of all sizes on applying modern machine learning techniques and is the creator of popular Python packages like StructureBoost, ML-Insights, and SplineCalib. He has also taught at UC-Berkeley, Brown University, and USF, bringing a unique mix of academic depth and real-world ML expertise.

Today, we explore how to make more reliable predictions in machine learning. From the dominance of gradient boosting for tabular data to the power of probabilistic regression and uncertainty quantification, Brian shares expert insights into building trustworthy ML models. We also dive into probability calibration, model drift, and best practices for ensuring model reliability in production.

Whether you're an ML engineer, data scientist, or business leader looking to improve your AI models, this episode is packed with practical takeaways you won’t want to miss.

 Key Topics Covered
✅ Gradient Boosting vs. Deep Learning – Why decision trees still dominate tabular data and structured business problems.
✅ Probabilistic Regression – Moving beyond point estimates to provide probability distributions and confidence intervals.
✅ Uncertainty Quantification – Understanding the limits of machine learning predictions and why it matters.
✅ Probability Calibration – How to ensure your model’s confidence scores are truly reliable.
✅ Handling Model Drift – Strategies to maintain model performance in a changing world.
✅ Real-World Use Cases – Applications in finance, healthcare, risk modeling, and business decision-making.

Resources & Tools Mentioned
Brian’s Youtube Channel: https://www.youtube.com/c/numeristical
Brian’s Linkedin: https://www.linkedin.com/in/brianlucena/
🛠️ StructureBoost – Brian’s open-source package for structured categorical variables in gradient boosting: https://github.com/numeristical/structureboost
📦 ML-Insights – Tools for better understanding ML models: https://github.com/numeristical/introspective
SplineCalib – A library for improving probability calibration: 🔗https://github.com/numeristical/splinecalib
📌 NGBoost – A gradient boosting approach for probabilistic regression: https://stanfordmlgroup.github.io/projects/ngboost/
🔗 GitHub for NGBoost: https://github.com/stanfordmlgroup/ngboost
📌 XGBoost – A powerful gradient boosting framework: https://github.com/dmlc/xgboost
📌 CatBoost – Gradient boosting with native support for categorical features: https://github.com/catboost/catboost
📌 LightGBM – A fast, efficient gradient boosting library: https://github.com/microsoft/LightGBM
📊 PyMC – A Bayesian probabilistic programming library for uncertainty modeling: https://github.com/pymc-devs/pymc

 Memorable Quotes
💬 "Businesses don’t just want a number—they need to understand a range of possible outcomes. That’s where probabilistic regression makes all the difference."
💬 "One of the biggest challenges in real-world ML is that the world doesn’t stay the same—models can drift, and retraining isn’t always the best solution."
💬 "Gradient boosting still outperforms deep learning for structured data because it handles sharp decision boundaries better."

This episode was sponsored by:
🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/
Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.


How to Make More Reliable Predictions in Machine Learning with Brian Lucena

In this episode of ODSC’s Ai X Podcast, The AI-Powered Analyst: Skills You Need to Stay Relevant, we explore how AI is revolutionizing analytics, decision-making, and career paths.

Our expert guests bring insights from both industry and academia:
 🔹 Dr. Aleksandar (Sasha) Tomic, is the Associate Dean for Strategy, Innovation & Technology at Boston College, oversees graduate programs in Applied Analytics and Applied Economics.
 🔹 Gary Arora, is a Chief Architect for Cloud & AI Solutions at Deloitte, Adjunct Professor at Boston College, and open-source contributor with over 5 million project downloads.

Topics Covered
🔹 AI’s impact on analytics and automation – How AI is transforming traditional data workflows
 🔹 The shift from SQL-based analysis to AI-powered decision-making – Can business users now replace data scientists?
 🔹 Challenges in enterprise AI adoption – Trust, governance, and data integration
 🔹 Agentic AI and autonomous decision-making – Are we replacing analysts or augmenting them?
 🔹 Essential skills for the AI-powered analyst – What professionals need to learn to future-proof their careers
 🔹 AI in education – How universities are adapting analytics programs to teach AI literacy
 🔹 The evolving job market – New roles emerging in AI-driven analytics
 🔹 Knowledge graphs in AI – How enterprises are integrating them into AI workflows
 🔹 Data quality and governance – The critical role of data integrity in AI-powered insights

Memorable Quotes
💬 "We’re not replacing the pilot, we’re building a better cockpit." – Gary Arora
 💬 "The premium for the human in the AI loop is not writing code—it’s knowing when things don’t seem right." – Sasha Tomic
 💬 "Your industry expertise will be your competitive advantage over AI." – Gary Arora
 💬 "Knowing when AI is wrong and why is now a critical skill for analysts." – Sheamus

This episode was sponsored by:
🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/
Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.
🔥 Use the exclusive code ODSCPodcast for an additional 10% off any ticket type!


The AI-Powered Analyst: Skills You Need to Stay Relevant

In this episode, we welcome Ivan Lee, founder and CEO of Datasaur.ai, a leading expert in LLM and AI solutions, particularly focusing on large language and reasoning models. Today, we delve deep into reasoning models, their recent advancements, real-world use cases, and the practical considerations enterprises should be aware of when adopting these powerful tools.
Whether you're a data scientist, ML engineer, practitioner, or business leader, this conversation offers valuable insights into one of AI’s most dynamic and impactful areas.

Key Topics Covered
Understanding Reasoning Models – How they differ from traditional large language models by explicitly demonstrating a chain of thought. 
Recent Developments – Exploring breakthroughs like OpenAI’s O1 and O3-mini, DeepSeek R1, Anthropic’s Claude 3.7 Sonnet, Google’s Gemini Flash Thinking, and Alibaba’s Qwen-32B. 
Real-World Applications – Where reasoning models excel, such as strategic research, complex analytics, and nuanced decision-making scenarios.
Challenges and Limitations – Addressing cost, latency, complexity, and non-determinism in practical deployments. 
Fine-Tuning Reasoning Models – Current challenges, techniques, and the potential of closed-beta programs. 
Agentic Workflows – Using reasoning models within agentic systems, their strengths, weaknesses, and realistic deployment considerations.
Future of Reasoning Models – The shift towards specialized, domain-specific models and ongoing commoditization of technology.

Memorable Quotes
💬 "These reasoning models unlock a whole new set of use cases that we didn't believe traditional foundation models would be able to address previously. But the downside is cost and latency”
💬 "A couple of weeks ago, we saw the releases of both Cloud 3.7 as well as OpenAI's GPT 4.5. And what was important about those announcements was as much about what they offered as what they didn't offer. 
💬 "I think a lot of people have been very hyped for a long time to see OpenAI's 4.5 model. Now that we got our hands on it, it really only demonstrated an incremental improvement on what we've seen previously.”
💬 "What could we accomplish if the model is really thinking hard about a problem for longer durations”

Resources & Tools Mentioned
Datasaur.ai Platform: www.datasaur.ai
Ivan Lee on LinkedIn: linkedin.com/in/ivanlee
Anthropic’s Claude 3.7 Sonnet: https://www.anthropic.com/news/claude-3-7-sonnet
OpenAI’s O1 and O3-mini: https://openai.com/index/openai-o3-mini/
Open AI’s Operator: https://openai.com/index/introducing-operator/
DeepSeek R1: deepseek.com
Google's Gemini Flash Thinking: https://deepmind.google/technologies/gemini/flash-thinking/
Alibaba's Qwen-32B Model: https://qwenlm.github.io/blog/qwq-32b/
"Thinking, Fast and Slow" by Daniel Kahneman: https://en.wikipedia.org/wiki/Thinking,_Fast_and_Slow
Manus AI - Universal AI Agent platform mentioned in the context of agentic workflows: https://www.manus.ai
OpenAI Operator (Agentic system for web-based tasks: https://openai.com/blog/operator
Deep Research by OpenAI: https://openai.com/blog/deep-research
Tencent's Huanyuan Turbo S - a new model allowing fast and slow thinking: https://ai.tencent.com
Firecrawl AI - Headless browser tool for data collection:  https://github.com/mendableai/firecrawl
Browserbase - another headless browser tool explicitly: https://browserbase.com
DeepSeek Open Source Week: https://deepseek.com/blog

This episode was sponsored by:
🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/
Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.
🔥 Use the exclusive code ODSCPodcast for an additional 10% off any ticket type!


Reasoning Models: Practical Insights from Ivan Lee

In this episode, we dive into how generative AI is shifting from experimentation to real product development—and why it’s no longer just for machine learning experts. Our guest is Hugo Bowne-Anderson, a well-known voice in the data and AI world. Hugo has spent years at the intersection of education, developer experience, and AI product thinking.
We explore how everyone—from software engineers and data scientists to product managers and designers—now has a seat at the table when it comes to building with AI. We also dig into why now is the time to move beyond “prompt and pray” and start creating applications people actually want to use.
Whether you’re hands-on technical or just getting started, this conversation will give you both insight and inspiration.
Key Topics Covered
Why now is the right time for a broader range of people to start building with generative AI
The shift from “prompt and pray” to real product development
Empowering people beyond ML experts—PMs, designers, SWE, DS—to participate meaningfully
Lowering barriers to entry with new tools (e.g. Cursor, Gemini, Replit)
Building LLMs, Agentic AI, and RAG Systems: Chunking, JSON Outputs, Latency
“Explorers and Settlers” Framework for AI Teams
The importance of evaluation-driven development for iterating from prototypes to real-world apps
How AI product development is becoming more collaborative and inclusive
Final Thoughts: What to Build Next and Why You Should

Memorable Outtakes:
💬 We’re no longer in the ‘prompt and pray’ era. The bar is higher—but the opportunity is bigger too.”
💬 “ AI is not gonna replace you. People who do what you do using AI will replace you.”
💬 “Everyone can build with AI—not just engineers. PMs, designers, analysts… they all shape the product.”

🔗 References & Resources
- Hugo’s Upcoming course, Building LLM-Powered Software for Data Scientists and Software Engineers: https://maven.com/hugo-stefan/building-llm-apps-ds-and-swe-from-first-principles?promoCode=ODSC
- Use the code ODSC for an additional 10% off.
- Hugo’s LinkedIn: https://www.linkedin.com/in/hugo-bowne-anderson-045939a5/
- Hugo’s Twitter: https://x.com/hugobowne
- Hugo’s Substack Newsletter: https://hugobowne.substack.com/
- Systematic Survey of Prompt Engineering: https://arxiv.org/pdf/2406.06608
- O'Reilly Radar essay, Escaping POC Purgatory: Evaluation-Driven Development for AI Systems: https://www.oreilly.com/radar/escaping-poc-purgatory-evaluation-driven-development-for-ai-systems/
- Essay -  Beyond Prompt-and-Pray: Building Reliable LLM-Powered Software in an Agentic World: https://www.oreilly.com/radar/beyond-prompt-and-pray/
- Build Reliable LLM-Powered Apps with Evaluation-Driven Development, Hugo’s free 10-day email course: https://vanishing-gradients.kit.com/8fff155b0c
- The Vanishing Gradients podcast: https://vanishinggradients.fireside.fm/
- High Signal: https://high-signal.delphina.ai/

This episode was sponsored by:
🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/
Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.
🔥 Use the exclusive code ODSCPodcast for an additional 10% off any ticket type!

Beyond Prompt and Pray: Why Now is the Time for Everyone to Build with AI

In this episode, we explore why graph technology is becoming foundational to the next wave of AI—especially as systems grow more complex and context becomes essential. Our guest is Amy Hodler, founder of GraphGeeks and a longtime leader in connected data, AI strategy, and graph analytics.
Amy takes us on a deep dive into how graphs offer a more natural, human-centered way to model data—focusing not just on what things are, but how they relate. We talk about GraphRAG, the blend of graph structures with retrieval-augmented generation, and how it’s giving AI systems a richer, more contextual backbone.
Whether you're an ML engineer, a data scientist, or just graph-curious, this conversation is full of insight into why relationships are the key to relevance in AI.

Key Topics Covered:
Why graphs are a natural fit for how humans think and make decisions

The evolution of GraphRAG and why it matters for building smarter LLM systems

How context-rich graphs improve AI explainability, fairness, and performance

Where graphs are gaining traction: cybersecurity, legal review, systems engineering

The growing importance of entity resolution and building high-quality graphs

Best practices for getting started with GraphRAG—and common pitfalls to avoid

Emerging trends: multimodal graph systems, agentic AI workflows, and graph-vectors hybrids

Why AI can’t solve every problem—and what to watch out for in the age of “AI snake oil”

Memorable Outtakes
 💬 “Graphs aren’t just about data points. They’re about what makes those points meaningful—relationships.”
 💬 “If you care about context, you should care about graphs.”
 💬 “GraphRAG isn’t just another buzzword. It’s how we bring structure and context into generation.”
 💬 “AI without context is guessing. Graphs help it make sense.”


🔗 References & Resources
- Amy’s LinkedIn: https://www.linkedin.com/in/amyhodler/
- Amy’s Twitter/X: https://x.com/amyhodler
- GraphGeeks’ Website: https://www.graphgeeks.org/
- GraphGeeks’ LinkedIn: https://www.linkedin.com/company/graphgeeks/
- Graph Algorithms: Practical Examples in Apache Spark and Neo4j: https://www.amazon.com/Graph-Algorithms-Practical-Examples-Apache/dp/1492047686
- David Hughes’ LinkedIn: https://www.linkedin.com/in/dahugh/overlay/about-this-profile/
- Amy’s ODSC East 2025 Session, Advancing GraphRAG: Text, Images, and Audio for Multimodal Intelligence: https://odsc.com/speakers/advancing-graphrag-text-images-and-audio-for-multimodal-intelligence/
- Blog, What is GraphRAG? An In-Depth Look at This Graph-Based Tool: https://opendatascience.com/what-is-graphrag-an-in-depth-look-at-this-graph-based-tool/
- AI Snake Oil: What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference: https://www.amazon.com/Snake-Oil-Artificial-Intelligence-Difference/dp/069124913X


This episode was sponsored by:
🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/
Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.
🔥 Use the exclusive code ODSCPodcast for an additional 10% off any ticket type!


Making AI Make Sense with Graphs: Context, Connections, and GraphRAG with Amy Hodler

In this episode, we speak with Valentina Alto, a Technical Architect specializing in AI & Apps at Microsoft, about the transformative potential of AI Agents. Valentina shares her insights into why major enterprises are increasingly interested in adopting AI Agent technology, highlighting its role in reshaping automation, enhancing productivity, and redefining customer experiences. With her deep expertise in generative AI, large language models (LLMs), and multi-agent systems, Valentina provides valuable perspectives on how AI Agents differ fundamentally from traditional AI Assistants.

Key Topics Covered:
- Definition and key differentiators of AI Agents versus traditional AI Assistants
- Core characteristics of effective AI Agents, including autonomy, specialization, and actionable capabilities
- Importance of LLMs, multimodal models, and advancements in reasoning capabilities
- Real-world applications of AI Agents in industries such as manufacturing, pharmaceuticals, and customer service
- Challenges such as error propagation and alignment of AI Agent autonomy with user intent
- Best practices for AI alignment, prompt engineering, and domain-specific knowledge integration
- Practical guidance on the lifecycle, development, and enterprise deployment of AI Agents
- The future ecosystem of AI Agents and the frameworks necessary for scalable enterprise implementation

Memorable Outtakes:
💬 "Autonomy, tools, and a clear system message are the main ingredients that define an AI Agent."
💬 "Prompt engineering isn't just important, it’s essential—especially for developers shaping what an AI Agent is meant to accomplish."
💬 "The future of AI Agents will be about defining robust ecosystems where these agents can effectively collaborate and scale."

References & Resources:
- Valentina Alto’s LinkedIn: https://www.linkedin.com/in/valentina-alto-6a0590148/recent-activity/all/
- Valentina Alto’s Notion Articles: https://shine-snowman-27f.notion.site/My-Articles-4f2f01923575438ba28e7a78ffb494a1
- Valentina Alto’s GitHub: https://github.com/valentina-alto
- Books by Valentina Alto:
   - Modern Generative AI with ChatGPT and OpenAI Models: https://www.packtpub.com/en-us/product/modern-generative-ai-with-chatgpt-and-openai-models-9781805123330
   - Building LLM Powered Applications: https://www.packtpub.com/product/building-llm-powered-applications/9781835462317
   - ODSC East Conference: https://odsc.com/boston/

Sponsored by: 🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/ Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.
🔥 Use the exclusive code ODSCPodcast for an additional 10% off any ticket type!


AI Agents: The Shift from AI Assistants to Intelligent Automation with Valentina Alto

In this episode of the ODSC Podcast, we sit down with Sid Probstein, a seasoned enterprise technologist, 10-time CTO, and now CEO of SWIRL. Sid shares deep insights on the evolving role of AI search in the enterprise, the limitations of current AI architectures like RAG, and how federated search powered by large language models (LLMs) can offer a more scalable, secure, and efficient approach.
This wide-ranging conversation also touches on the future of agentic AI, the implications of zero-trust architecture, and why moving all your enterprise data into vector databases may not be the answer.
Key Topics Covered:
- Why many first-generation enterprise AI architectures fall short
- The common misconception that RAG requires vector databases
- How federated search is being reinvented with the help of LLMs
- The differences between RAG, AI search, and hybrid approaches
- Why data movement is a privacy, governance, and scaling bottleneck
- Using LLMs to evaluate, re-rank, and extract relevant results from existing search engines
- The impact of AI search on agentic workflows and enterprise automation
- Architectural considerations for building secure, scalable AI systems
- How open-source tools like Swirl can jumpstart your AI search strategy
- The future of applications and employment in an AI agent-driven world
Memorable Outtakes:
💬 "A lot of those first-generation architectures are just dead wrong."
 – Sid Probstein on the pitfalls of blindly applying RAG and vector database stacks in enterprise environments.
💬 "You do not need a vector database to do RAG. The idea that RAG is owned by the vector databases is absurd."
 – Sid debunks a common myth about RAG and data pipelines.
💬 "Federated search was niche—until LLMs changed everything."
 – Sid on how modern LLMs make sense of heterogeneous data and breathe new life into federated search.
References & Resources:
- Sid Probstein: https://www.linkedin.com/in/sidprobstein/
- Swirl (Open Source Project): https://github.com/swirlai/swirl-search
- Swirl Website: https://www.swirl.today
- Wired Article on COBOL Dates & Government Data (referenced in episode): https://www.wired.com/story/ai-small-business-administration-loans-errors/
- Estonia’s Digital Government Model (mentioned): https://e-estonia.com
- LangChain (for prompt engineering and agent workflows): https://www.langchain.com
- Open Source LLM Deployment with Ollama: https://ollama.com
- Deepseek R1:  https://github.com/deepseek-ai/DeepSeek-R1
Sponsored by:
This episode was sponsored by:
 🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/
 Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.
🔥 Use the exclusive code ODSCPodcast for an additional 10% off any ticket type!

Rethinking RAG: Why AI Search Needs a New Architecture with Sid Probstein

In this episode, we explore the frontiers of AI research with Eric Xing, President of Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), Professor at Carnegie Mellon University, and founder of GenBio AI. A pioneer in AI Biology, Eric discusses the evolving landscape of world models, agentic AI, and the revolutionary concept of AI-driven digital organisms. With a background that spans computational biology, large-scale ML systems, and foundational research, Eric provides deep insights into how AI is transforming our understanding of intelligence, life, and scientific simulation.

Key Topics Covered
Defining World Models:
-What they are, how they differ from generative models, and why they are essential for reasoning and simulation.

The Role of World Models in Agentic AI:
-How simulation and planning rely on internal representations of environments.

Why Digital Agents Still Need World Models:
-Even in digital workflows, world models help with decision-making, abstraction, and tool interaction.

Agent Models and Autonomous Reasoning:
-The limitations of step-by-step optimization and the need for scenario simulation.

AI Biology and GenBio AI’s Vision:
-Introducing the concept of the AI-Driven Digital Organism (AIDO).
-Multiscale foundation models for DNA, RNA, proteins, and cells.
-Simulating biology to replace expensive and slow wet-lab experimentation.

Challenges in Modeling Life:
-Why LLM-style architectures don’t work for biology.
-The need for multimodal, co-trained, interpretable architectures.
-Explainability in AI Biology: How modular architectures and simulation-based evaluation can help bridge the gap.

Ethical Boundaries and Societal Readiness:
-Why AI progress is bounded by what society is ready and willing to adopt.

Memorable Outtakes
“World models are simulators of all possibilities—not just in the physical world, but also in mental and cyber spaces.”
 – Eric Xing, on expanding the scope of simulation in AI.
“We want to build a virtual cell—or an AI-driven digital organism—that can simulate biology at any scale.”
– Eric Xing, explaining the core vision of AI Biology and how it's redefining biological modeling.
“In biology, the research is still done in the wet lab. We want to replace that with a digital system that simulates all plausible hypotheses.”
– Eric Xing, on the need to digitize the discovery phase in medicine and drug development.

References & Resources Mentioned
Eric Xing on LinkedIn: https://www.linkedin.com/in/eric-xing-a6900418/
GenBio AI: https://genbio.ai
MBZUAI (Mohamed bin Zayed University of Artificial Intelligence): https://mbzuai.ac.ae
DeepMind Gemini Robotics: https://deepmind.google/technologies/gemini
Broad Institute: https://www.broadinstitute.org
David Ha & Jürgen Schmidhuber “World Models” paper: https://arxiv.org/abs/1803.10122
Fei-Fei Li’s World Labs: https://www.technologyreview.com/2024/03/12/1170123/fei-fei-li-world-models-startup/
Eric’s ODSC East Keynote Session, “Toward Public and Reproducible Foundation Models Beyond Lingual Intelligence”: https://odsc.com/speakers/toward-public-and-reproducible-foundation-models-beyond-lingual-intelligence/
Sponsored by

This episode was sponsored by:
 🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/
 Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.
🔥 Use the exclusive code ODSCPodcast for an additional 10% off any ticket type!

Can AI Simulate Life? Exploring World Models and Digital Organisms with Eric Xing

In this episode, we sit down with Thomas Wiecki, co-creator of PyMC and CEO of PyMC Labs, to explore the evolving frontier of probabilistic AI. With roots in computational psychiatry and a passion for Bayesian modeling, Thomas walks us through how probabilistic methods unlock better decision-making by quantifying uncertainty—something traditional AI often glosses over.
We dive deep into real-world applications, from marketing mix modeling to synthetic consumers, and hear about Thomas's upcoming session at ODSC East. Whether you're a marketer, data scientist, or just AI-curious, this episode provides an illuminating perspective on building interpretable, trustworthy AI systems.

Key Topics Covered
Defining Probabilistic AI:
What distinguishes probabilistic AI from traditional machine learning and generative models.
Why uncertainty is a feature, not a bug—and how it enables better risk-aware decisions.

Marketing Mix Modeling (MMM):
How PyMC Labs applies Bayesian models to measure marketing effectiveness across channels.
Real-world examples of ROI estimation, saturation detection, and spend optimization.

AI Agents and Natural Language Interfaces:
How PyMC Labs built AI agents to interpret and interact with complex statistical models.
Bridging the gap between marketers and data scientists through conversational AI.

Synthetic Consumers and Product Testing:
Using LLMs to simulate consumer feedback and behavior before market release.
How these systems complement surveys and reduce time-to-insight.

Building Trustworthy AI:
Why transparency and causal reasoning are critical for adoption in business.
The role of PyMC in making statistical modeling accessible, powerful, and interpretable.

Community and Open Source:
The evolution of the PyMC ecosystem and the role of its global contributor base.
How open-source collaboration leads to robust, production-ready tools.

Scalability and Next-Gen Tooling:
Leveraging GPUs, Apple Silicon, and frameworks like Aesara, JAX, and MLX to scale inference.
The future of PyMC agents and integrations with platforms like Databricks and MLFlow.

Memorable Outtakes
“Probabilistic AI doesn’t just make predictions—it tells you how confident it is. That’s a game-changer for real-world decisions.”
“The agent can now say, ‘I'm not sure about this prediction—you might want to collect more data.’ That’s the kind of reasoning we need from AI.”
“Synthetic consumers sound like science fiction, but they’re a practical tool for product development today.”
“We’ve moved from dashboards to conversations. Marketers can now ask a question—and get a data-driven, contextual answer back.”
“Open source isn’t just about free tools. It’s about building the best ideas with the best minds from around the world.”

Resources:
- Thomas’s LinkedIn: https://www.linkedin.com/in/twiecki/
- Thomas’s Website: https://www.pymc-labs.com/
- Thomas’s GitHub: https://github.com/twiecki
- PyMC: https://www.pymc.io/welcome.html
- PyMC-Labs: https://www.pymc-labs.com/
- PyMC GitHub: https://github.com/pymc-devs/pymc
- Marketing Mix Modeling Agent: https://www.pymc-labs.com/blog-posts/the-ai-mmm-agent/
- Thomas’s ODSC East session, “Automating Bayesian Marketing Mix Modeling with an AI-Powered Assistant,”: https://odsc.com/speakers/automating-bayesian-marketing-mix-modeling-with-an-ai-powered-assistant/
- Allen Downey’s (also with PyMC) ODSC East session, “Mastering Time Series Analysis with StatsModels: From Decomposition to ARIMA,”: https://odsc.com/speakers/mastering-time-series-analysis-with-statsmodels-from-decomposition-to-arima/

Sponsored by
This episode was sponsored by:
 🎤 ODSC East 2025 – The Leading AI Builders Conference – https://odsc.com/boston/
 Join us from May 13th to 15th in Boston for hands-on workshops, training sessions, and cutting-edge AI talks covering generative AI, LLMOps, and AI-driven automation.
🔥 Use the exclusive code ODSCPodcast for an additional 10% off any ticket type!

Inside Probabilistic AI: Bayesian Modeling and PyMC with Thomas Wiecki

In this episode of the ODSC AIX Podcast, host Sheamus McGovern speaks with Arvind Narayanan, professor of computer science at Princeton University and co-author of the widely acclaimed book AI Snake Oil. Arvind is also the director of the Center for Information Technology Policy and one of TIME's 100 most influential people in AI.
Together, they unpack the growing disconnect between AI hype and practical deployment, focusing particularly on agentic AI systems and why current evaluation methods are falling short. Arvind discusses the moral limits of predictive AI, the myth of superintelligence, and how practitioners can ground their optimism in accountability, transparency, and more realistic expectations.

Key Topics Covered
- The origins and impact of AI Snake Oil
- Why many AI systems are overhyped or misapplied
- Predictive vs. generative AI: what's really at stake
- Why superintelligence is a myth—and a distraction
- The flawed assumption behind “upstream” evaluations
- Engineering and ethical limitations of AI agents in real-world deployment
- Misinformation, misuse, and human responsibility in the AI age
- The importance of resilience over restriction in AI governance

Memorable Outtakes
💬 “We came to the conclusion that superintelligence is a bit of a myth. AI will continue to improve, but …”
💬 “Broken AI is often appealing to broken institutions. You can’t fix a flawed hiring process with more automation—you need more humanity.”
💬 "Upstream evaluation is mostly dead."

References & Resources
- Arvind Narayanan Academic Profile: https://www.cs.princeton.edu/~arvindn/
- AI Snake Oil (Book & Newsletter)  https://www.aisnakeoil.com
- “How to Recognize AI Snake Oil” (Slides)
 https://www.cs.princeton.edu/~arvindn/talks/MIT-STS-AI-snakeoil.pdf
- AI Agents That Matter (Research Paper) https://arxiv.org/abs/2407.01502
- Holistic Agent Leaderboard (HAL)   https://hal.institutions.ai
- Essay: “AGI Is Not a Milestone”
 https://aisnakeoil.substack.com/p/agi-is-not-a-milestone
- Essay: “AI as Normal Technology”
 https://aisnakeoil.substack.com/p/ai-as-normal-technology
- HAL: Holistic Agent Leaderboard  https://hal.cs.princeton.edu/
- Agentic AI Summit
 https://www.summit.ai/
- ODSC West –  https://odsc.com/california
- Arvind’s ODSC East Keynote Session, AI Agents: Taking Stock and Looking Ahead: https://odsc.com/speakers/ai-agents-taking-stock-and-looking-ahead/

Sponsored by:
Agentic AI Summit 2025
 Join the premier virtual event for AI builders from July 15–30. Gain hands-on skills in designing, deploying, and scaling autonomous AI agents.
 🔥 Use code podcast for 10% off any ticket.
 Register now: https://www.summit.ai/
ODSC West 2025 – The Leading AI Training Conference
 Attend in San Francisco from October 28–30 for expert-led sessions on generative AI, LLMOps, and AI-driven automation.
 🔥 Use code podcast for 10% off any ticket.
 Learn more: https://odsc.com/california

The AI Superintelligence Myth with Arvind Narayanan

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.

Rethinking RAG: Why AI Search Needs a New Architecture with Sid Probstein

ODSC's Ai X Podcast

Related tracks