The highest-signal Python AI series
Meetups and conferences from Pydantic and Prefect—for Python engineers shipping AI to production, founders of breakout AI products, and technical leaders operating agents and systems at scale. Short talks, honest demos, and conversation that survives contact with production.
How Py AI started
It started with a walk through Manhattan. Samuel and Adam kept circling the same frustration: plenty of AI events existed, but almost none spoke to the people they kept meeting—Python developers trying to get AI systems to actually work in production.
That frustration shows up wherever teams ship workflows and agents: when reliability has to hold, iteration can't mean silent regressions, and the fires start the moment real users show up. The hard part was never getting a demo to run. It was everything after.
The hosts spend their days in that stack—orchestration, agents, MCP, observability—so when they say the hard part isn't the demo, it isn't a slogan. It's what lands in their inboxes.
Py AI exists to make room for those conversations: grounded in what it takes to ship, next to the Python data and AI tooling work the community already depends on.
Samuel Colvin
CEO, Pydantic
Adam Azzam
VP Product, Prefect
“We shipped something that looked great in a notebook, then spent six weeks making it not fall over.”
What we kept hearing
The pattern was everywhere. Smart teams, solid models, and then:
- Evals that didn't catch what mattered
- Feedback loops that took days instead of minutes
- Debugging distributed systems problems, not model problems
- No clear patterns for safety, observability, or code that doesn't become legacy the moment you look away
The layer Py AI cares about
A lot of production AI fails in the glue: pipelines that stall halfway, agents that don't recover, jobs that looked fine in a notebook and flake under load. Py AI tilts toward that layer—scheduling, durable execution, observability when the bug isn't in the model, MCP where security and ops get a real vote—not because keynotes are boring, but because that's where the audience is often stuck.
Sessions aren't roadmap theater. They're built so when someone says “what broke in prod,” the thread can stay on retries, backpressure, timeouts, and who owns side effects—long enough to turn into something you can try the next day.
What we're building
We curate sessions where you walk out with something to try tomorrow. That means clear stories, concrete tradeoffs, and details that hold up when you push back. No hype. No merchants of complexity. Just builders comparing notes.
What we're trying to foster
- A room full of people who share what they tried, what broke, and what they'd do differently
- Short talks and honest demos, then real conversation
- Practitioners, not evangelists
- Ideas you can actually use
How we gather
- Local gatherings: short talks, live demos, and plenty of time to meet the person next to you.
- Annual conference: curated talks plus an open call for proposals, because the best insights often come from unexpected places.
Previous events & what's next
Past meetups and speakers are on pyai.events/events; recordings on the Py AI YouTube channel. Schedules, registration, and the full program live on pyai.events—including why Py AI exists. For upcoming dates, subscribe on Luma.
Come find us
Subscribe for upcoming dates, or jump to the main site for everything in one place.