PyData London
@pydatalondon.bsky.social
https://london.pydata.org
We run a monthly meetup and host the annual PyData London conference. PyData is an educational program of NumFOCUS, helping our community share ideas and learn from each other.
https://www.meetup.com/PyData-London-Meetup/
We run a monthly meetup and host the annual PyData London conference. PyData is an educational program of NumFOCUS, helping our community share ideas and learn from each other.
https://www.meetup.com/PyData-London-Meetup/
1. And finally we have a lightning talk on USearch: The Engine Behind Most New Search & RAG Pipelines - Ash Vardanian
USearch might be the most widely used search engine you’ve never heard of, up there with FAISS and Lucene. It even supports semantic search and geospatial indexing!
USearch might be the most widely used search engine you’ve never heard of, up there with FAISS and Lucene. It even supports semantic search and geospatial indexing!
November 4, 2025 at 9:26 PM
1. And finally we have a lightning talk on USearch: The Engine Behind Most New Search & RAG Pipelines - Ash Vardanian
USearch might be the most widely used search engine you’ve never heard of, up there with FAISS and Lucene. It even supports semantic search and geospatial indexing!
USearch might be the most widely used search engine you’ve never heard of, up there with FAISS and Lucene. It even supports semantic search and geospatial indexing!
Our first lightning talk is Mise and easy: managing environments the fun way - Connor Adams. Environment management is very messy across languages. Mise can make your life much easier!
November 4, 2025 at 9:21 PM
Our first lightning talk is Mise and easy: managing environments the fun way - Connor Adams. Environment management is very messy across languages. Mise can make your life much easier!
November 4, 2025 at 8:42 PM
It works with CV, JSON, JSONL and YAML,
has custom deserialisation, many (many!) out of the box fixtures, and even parameterisation.
has custom deserialisation, many (many!) out of the box fixtures, and even parameterisation.
November 4, 2025 at 8:37 PM
It works with CV, JSON, JSONL and YAML,
has custom deserialisation, many (many!) out of the box fixtures, and even parameterisation.
has custom deserialisation, many (many!) out of the box fixtures, and even parameterisation.
Pytest fixtures great but are slightly strange things. They feel a bit too much like magic sometimes.
@feregri.no solution is pytest-fixture-fixture (really!)
Fixtures, for fixtures. Handy fixtures to access your test fixtures from your pytest tests.
@feregri.no solution is pytest-fixture-fixture (really!)
Fixtures, for fixtures. Handy fixtures to access your test fixtures from your pytest tests.
November 4, 2025 at 8:29 PM
Pytest fixtures great but are slightly strange things. They feel a bit too much like magic sometimes.
@feregri.no solution is pytest-fixture-fixture (really!)
Fixtures, for fixtures. Handy fixtures to access your test fixtures from your pytest tests.
@feregri.no solution is pytest-fixture-fixture (really!)
Fixtures, for fixtures. Handy fixtures to access your test fixtures from your pytest tests.
Data-driven testing is an approach where test logic stays the same, but the input data and expected outputs come from external sources.
Here’s a Pytest example.
Here’s a Pytest example.
November 4, 2025 at 8:24 PM
Data-driven testing is an approach where test logic stays the same, but the input data and expected outputs come from external sources.
Here’s a Pytest example.
Here’s a Pytest example.
Next tonight we’re hearing about No More Boilerplate: Data-Driven Testing Made Easy from @feregri.no
November 4, 2025 at 8:19 PM
Next tonight we’re hearing about No More Boilerplate: Data-Driven Testing Made Easy from @feregri.no
Agents mean we can store data collected in the agent state and use that to feel more comfortable in what they predict vs a single (overconfident) LLM.
But don’t add AI tools if you don’t have to! Deterministic code is simpler and more reliable.
But don’t add AI tools if you don’t have to! Deterministic code is simpler and more reliable.
November 4, 2025 at 7:45 PM
Agents mean we can store data collected in the agent state and use that to feel more comfortable in what they predict vs a single (overconfident) LLM.
But don’t add AI tools if you don’t have to! Deterministic code is simpler and more reliable.
But don’t add AI tools if you don’t have to! Deterministic code is simpler and more reliable.
There are many frameworks for creating agents but Geir is a fan of @hf.co smolagents. It’s simple to use and powerful enough for production.
November 4, 2025 at 7:41 PM
There are many frameworks for creating agents but Geir is a fan of @hf.co smolagents. It’s simple to use and powerful enough for production.
We’re entering the ALL CAPS era of doctstrings—because they end up in system prompts.
November 4, 2025 at 7:31 PM
We’re entering the ALL CAPS era of doctstrings—because they end up in system prompts.
Just asking an LLM to predict which team will win doesn’t work that well. It misinterprets data and doesn’t have enough domain knowledge to build trust and consistency.
November 4, 2025 at 7:28 PM
Just asking an LLM to predict which team will win doesn’t work that well. It misinterprets data and doesn’t have enough domain knowledge to build trust and consistency.
Starting tonight’s meet up with @geirfreysson.bsky.social
Using Hugging Face's smolagents framework to create a Fantasy Premier League AI Agent.
His main motivation was to beat his brother at FPL!
Using Hugging Face's smolagents framework to create a Fantasy Premier League AI Agent.
His main motivation was to beat his brother at FPL!
November 4, 2025 at 7:19 PM
Starting tonight’s meet up with @geirfreysson.bsky.social
Using Hugging Face's smolagents framework to create a Fantasy Premier League AI Agent.
His main motivation was to beat his brother at FPL!
Using Hugging Face's smolagents framework to create a Fantasy Premier League AI Agent.
His main motivation was to beat his brother at FPL!
Up next: @hevansdev.bsky.social Answering a dumb question with a smart bird feeder.
Is this robin abnormally large?
Is this robin abnormally large?
October 7, 2025 at 8:05 PM
Up next: @hevansdev.bsky.social Answering a dumb question with a smart bird feeder.
Is this robin abnormally large?
Is this robin abnormally large?
Now hearing HangMAN vs AI - Kavit Tolia
Learnings building a custom transformer to play hangman.
Learnings building a custom transformer to play hangman.
October 7, 2025 at 7:54 PM
Now hearing HangMAN vs AI - Kavit Tolia
Learnings building a custom transformer to play hangman.
Learnings building a custom transformer to play hangman.
Next up Why Python is the best language for algorithmic trading from @richardhickling.bsky.social
October 7, 2025 at 7:44 PM
Next up Why Python is the best language for algorithmic trading from @richardhickling.bsky.social
Now hearing from Victor Zommers on his experiments getting LLMs to play Go. They were not good, not only losing to KataGo but also Wally, developed in 1981 with just 7 rules.
October 7, 2025 at 7:03 PM
Now hearing from Victor Zommers on his experiments getting LLMs to play Go. They were not good, not only losing to KataGo but also Wally, developed in 1981 with just 7 rules.
Next up we have Sam Joseph LLMs can tell jokes, but they’re pretty bad. Sam found he can breed better jokes using ideas from evolutionary computing—with pretty decent results!
October 7, 2025 at 6:57 PM
Next up we have Sam Joseph LLMs can tell jokes, but they’re pretty bad. Sam found he can breed better jokes using ideas from evolutionary computing—with pretty decent results!
Our first talk tonight is from PyDatalondon founder @ianozsvald.bsky.social on his playgroup: a collaborative space to explore really difficult but interesting problems together like ARC AGI, diagnosing bugs agenically and making LLMs funny.
October 7, 2025 at 6:45 PM
Our first talk tonight is from PyDatalondon founder @ianozsvald.bsky.social on his playgroup: a collaborative space to explore really difficult but interesting problems together like ARC AGI, diagnosing bugs agenically and making LLMs funny.
From small beginnings to great things! @emlynclay.bsky.social tells us the history of PyData London.
October 7, 2025 at 6:28 PM
From small beginnings to great things! @emlynclay.bsky.social tells us the history of PyData London.
It’s finally time! 🥳 PyData London 100 is here! 💯
October 7, 2025 at 6:13 PM
It’s finally time! 🥳 PyData London 100 is here! 💯
You don’t need a degree to get into data science: apprenticeships are also great.
And there are many benefits. Empowered and loyal, apprentices drive plenty of fresh thinking with real world impact: 87% of employers say apprentices boost team innovation.
And there are many benefits. Empowered and loyal, apprentices drive plenty of fresh thinking with real world impact: 87% of employers say apprentices boost team innovation.
September 3, 2025 at 12:02 AM
You don’t need a degree to get into data science: apprenticeships are also great.
And there are many benefits. Empowered and loyal, apprentices drive plenty of fresh thinking with real world impact: 87% of employers say apprentices boost team innovation.
And there are many benefits. Empowered and loyal, apprentices drive plenty of fresh thinking with real world impact: 87% of employers say apprentices boost team innovation.
Last but definitely not least we have Ruby Waller on Apprenticeships in Data: a practical pathway into the profession.
September 2, 2025 at 11:52 PM
Last but definitely not least we have Ruby Waller on Apprenticeships in Data: a practical pathway into the profession.
There are many unresolved problems with agents.
Context engineering helps. Designing the entire informational environment not just the prompt so the model can "plausibly" complete the task.
Context engineering helps. Designing the entire informational environment not just the prompt so the model can "plausibly" complete the task.
September 2, 2025 at 11:49 PM
There are many unresolved problems with agents.
Context engineering helps. Designing the entire informational environment not just the prompt so the model can "plausibly" complete the task.
Context engineering helps. Designing the entire informational environment not just the prompt so the model can "plausibly" complete the task.
Our first lightning talk from Imamuddin Shaik on Reliable Agents: Context Engineering, Roles, and What 1M-Token Windows Don’t Fix.
September 2, 2025 at 11:47 PM
Our first lightning talk from Imamuddin Shaik on Reliable Agents: Context Engineering, Roles, and What 1M-Token Windows Don’t Fix.
One surprisingly effective way to evaluate outputs is using an LLM as a judge.
Observability is key. Focus on process not tools.
Observability is key. Focus on process not tools.
September 2, 2025 at 7:38 PM
One surprisingly effective way to evaluate outputs is using an LLM as a judge.
Observability is key. Focus on process not tools.
Observability is key. Focus on process not tools.