Last but not least one of the best branded projects I've seen

IBM has a coding agent! BOB! Looks pretty interesting

Interesting view of how parallelization increases efficiency (in gpu compute processes)

Key point - if you have low utilization on GPU the cost per token is very high but high utilization is cheaper per token. Balance that against the API cost from a provider

Celebrity Pydata member Sebastian Wallkötter presenting on token costs!

OpenRAG open source project for offloading the RAG pipeline building

Rag vs direct context where it works best

At IBMs hacking agents meetup. Some interesting talks coming up will post some of what I'm learning

Back at it again! PyData Boston + Cursor hackathon

Double feature by @fayshaw.bsky.social !

Popping in to Boston Python meetup for @fayshaw.bsky.social's talk on LASER for learning analysis github.com/laser-instit...

#odsc @allendowney.bsky.social difference in life expectancy over time between men and women

#odsc Pydata frequent flyer and great dude @allendowney.bsky.social doing his Bayesian best at ODSC

Using LLMs to better "model" how a piece of software works based on tests and contextual information #odsc

Analysis bottleneck - coding tools mean more code to review, and there's only so many reviewers #odsc

Briefly stopping by #ODSC - liked this statement that AGI isn't necessary for societal improvements

Open source foundation model for time series - MOMENT github.com/moment-times...

Just what is time series classification?

#pydata our next speakers Abhishek Murthy and Evans Addo about time series analysis

#pydatabos setting up an experiment to see how LLMs treat speaker transitions

#pydatabos interesting diagram for explaining transformers poloclub.github.io/transformer-...

#pydatabos LLMs vs humans as participants in studies

#pydatabos first up is Julia Mertens on LLMs hallucinating speaker transitions

#pydatabos back at it for April!

#pydatabos different flavors of graphs

#pydatabos our second speaker on sparql-tool how to have your agent use knowledge graphs

#pydatabos interesting use of color to draw attention to specific parts of the visual

#pydatabos two different ways to use the gofish syntax

#pydatabos Grouping principles behind the gofish visualization library

#pydatabos visualization of this "shrinking" trend is easy to miss depending on the choice of visual