Finding AI insights and education at PyCon US 2026
While AI content is sprinkled throughout the event, how could it not be, PyCon US features a dedicated The Future of AI with Python track, new this year, and programmed by Elaine Wong, PyCon US Chair, Jon Banafato, PyCon US Co-Chair, and Philip Gagnon, Program Committee Chair. According to JetBrains' State of Developer Ecosystem 2025 report, 85% of developers now regularly use AI tools for coding and development (which tell us that you are probably doing that), and 62% rely on at least one AI coding assistant, agent, or code editor. Looking ahead, nearly half of all developers (49%) plan to try AI coding agents in the coming year, and the eight sessions in this track map onto those priorities, covering everything from running LLMs on your laptop to building real-time voice agents. T a look at the big themes and the sessions and tutorials you won't want to miss in our new track and throughout the event.
Let’s start with newbies, if you or someone on your team is just getting started with ML, Corey Wade's Wednesday tutorial Your First Machine Learning Models: How to Build Them in Scikit-learn is the perfect entry point, a hands-on introduction to the building blocks that underpin so much of what's discussed in the talks.
LLMs Are Moving to the Edge
One of the most significant shifts in AI right now is the move toward running models locally on laptops, browsers, and devices, rather than in centralized cloud infrastructure. Want to know more? Check out: Running Large Language Models on Laptops: Practical Quantization Techniques in Python from Aayush Kumar JVS, a hands-on look at how quantization makes large models practical on consumer hardware. Fabio Pliger takes a look at the role of the browser with Distributing AI with Python in the Browser: Edge Inference and Flexibility Without Infrastructure, exploring how Python-powered inference can run client-side with no server required. If you've been watching the open-weights model explosion and wondering how to actually deploy these things, these two talks are for you.
Want to go deeper before the conference even starts? On Wednesday, May 13th, Isabel Michel's tutorial Implementing RAG in Python: Build a Retrieval-Augmented Generation System gives you hands-on experience building a retrieval-augmented generation pipeline from scratch, the practical foundation underneath a lot of modern LLM applications.
AI Agents and Async Python
Agentic AI, systems that take multi-step actions autonomously, is one of the defining developments of 2025 and continues to take the world by storm in 2026. But building agents that actually work in production requires getting async right. Aditya Mehra's Don't Block the Loop: Python Async Patterns for AI Agents digs into the concurrency pitfalls that trip up so many teams when they move from demo to deployment. This talk bridges a gap that many tutorials leave open: the gap between "I have a working agent" and "my agent works reliably at scale."
If you want a running start, Pamela Fox's Wednesday tutorial Build Your First MCP Server in Python is the perfect on-ramp, MCP (Model Context Protocol) is quickly becoming the standard way to give AI agents access to tools and data, and building one yourself is the fastest way to understand how agentic systems actually work under the hood.
AI and Open Source Sustainability
AI-Assisted Contributions and Maintainer Load by Paolo Melchiorre tackles a genuinely thorny question: as AI tools make it easier to generate pull requests, what happens to the maintainers on the receiving end? Drawing on real examples from projects like GNOME, OCaml, Python, and Django, Melchiorre examines how AI-generated contributions are shifting workload onto already time-constrained maintainers and what the open source community is doing about it.
High-Performance Inference in Python
Python performance engineering is no longer optional for AI workloads. Yineng Zhang's High-Performance LLM Inference in Pure Python with PyTorch Custom Ops walks through the techniques for squeezing real speed out of inference pipelines without leaving the Python ecosystem (This one isn’t in the track, but it is so on point, I had to add it). Paired with Santosh Appachu Devanira Poovaiah's What Python Developers Need to Know About Hardware: A Practical Guide to GPU Memory, Kernel Scheduling, and Execution Models, Friday's track offers a practical hardware-to-application view of the performance stack that's increasingly essential for anyone building production AI systems.
Also, Catherine Nelson and Robert Masson's Thursday tutorial Going from Notebooks to Production Code is a great complement; bridging the gap between exploratory AI work and the kind of reliable, maintainable code that actually makes it into production systems.
Explainability and Responsible AI
As AI systems make more consequential decisions, the demand for explainability is only growing from regulators, from users, and from the developers building these systems. Jyoti Yadav's Building AI That Explains Itself: Why Your Card Got Declined uses a familiar real-world example to demonstrate how Python developers can build transparency into AI-driven decisions. It's a topic at the heart of current conversations about AI trust, and one that every practitioner should be thinking about.
Two tutorials round this theme out nicely: Neha's Wednesday session Causal Inference with Python teaches you how to move beyond correlation and reason about cause and effect in your data, a foundational skill for anyone building AI systems that need to explain why they made a decision. And on Thursday, Juliana Ferreira Alves' When KPIs Go Weird: Anomaly Detection with Python gives you practical tools for catching when your AI-powered systems go off the rails before your users do.
Voice AI and Multimodal Interfaces
Real-time voice is one of the fastest-moving areas in applied AI, and Camila Hinojosa Añez and Elizabeth Fuentes close out the AI track Friday evening with How to Build Your First Real-Time Voice Agent in Python (Without Losing Your Mind). This practical session covers the building blocks of voice agents in Python, a skill set that's quickly becoming table stakes for developers building consumer-facing AI products.
AI and Education
Sonny Mupfuni's AI-Powered Python Education: Towards Adaptive and Inclusive Learning explores how Python can power learning that adapts to the student, and Gift Ojeabulu's Making African Languages Visible: A Python-Based Guide to Low-Resource Language ID takes on one of NLP's most persistent blind spots, the languages that dominant datasets routinely leave out. Don't skip these. These sessions represent Python's role not just in building AI products, but in democratizing access to AI's benefits.
Friday's AI track is a rare chance to hear from practitioners who are building real things in production, not just demoing prototypes. Whether you're a Python developer who's been watching the AI wave from the sidelines or a team already shipping AI features who wants to sharpen your craft, clear your schedule and pull up a chair. And a big THANK YOU to Anaconda and NVIDIA for sponsoring this track!
We'll see you in Long Beach.
PyCon US 2026 takes place May 13–19 in Long Beach, California. The Future of AI with Python talk track runs Friday, May 15th.
Comments