The evening
- 18:00
Doors open
Arrive, grab food, test your laptop on the screen.
- 18:30
Keynote: Samuel Colvin
Creator of Pydantic, now running the company behind Pydantic AI and Logfire.
- ~19:00
Your talk
20 minutes. One speaker slot, so the room is yours.
- ~19:20
Break
Food, drinks, and a chance to catch the speakers.
- ~19:40
Discussion tables
Small-group conversations on what came up in the talks, until about 20:15.
Talks are in English. Live demos are welcome but not required.
Who's in the room
Mostly engineers shipping LLM products, plus product people and researchers. Pitch the talk at people who build these systems day to day.
The one rule
No sales pitches. Mention where you work and what you build, but the talk is about what you learned: what broke in production and how you fixed it.
Talks we'd love to hear
Case study
How you evaluate a specific RAG system, agent or LLM feature in your product: the metrics, the pipeline, and the hardest problems you hit.
Tooling deep dive
Frameworks, internal tools or infrastructure for evaluating, testing, monitoring or logging AI systems, and how you wired them into CI/CD.
Research & experiments
Papers, internal reports or experiments with practical consequences for people building production systems.
Open problems
Where current tools fall short: evaluating without ground truth, measuring hallucinations, judging agent behaviour. You don't need the answer.
Venue & AV
Want to speak?
Send a short note and we'll get back to you. Happy to talk an idea through before you commit.
Not sure your topic fits? Write anyway.
Where this comes from
Evals London is the sister meetup of Evals.cz in Prague, run by Šimon Podhajský and Veronika Pešková. Prague has held four editions since February 2026, and each one filled up with a waitlist.
Prague editions
- #1 3 Feb 2026 at Apify
- #2 15 Apr 2026 at Shoptet
- #3 17 Jun 2026 at Make
- #4 1 Oct 2026 at Truhlárna (Miton)
Watch edition #4
A joint evening with Miton AI Times: three talks, all on evals. Full recording on YouTube
- The Weird and Wonderful World of Cybersecurity Evals
- Evaluating Open Models on Your Task and in Your Language
- Generating Synthetic Evaluation Data for Search Agents