Boundary ML
@boundaryML
Developing BAML, the AI framework that puts the engineering back into prompt engineering.
You might like
We decided to make our function-calling benchmark results fully public and interactive!
Our interactive function-calling benchmark results are live! boundaryml.com/blog/sota-func…
boundaryml.com
Beating OpenAI's structured outputs on cost, accuracy and speed — An interactive deep-dive
We leveraged a novel technique, schema-aligned parsing, to achieve SOTA on BFCL with every LLM.
Amazing combo of #opensource @cocoindex_io and @boundaryML to build resilient, live structured extraction in LLM data flows. Fully open source, link in comment!
Only in SF kinda story. Saw a beautiful girl at this cafe I was working from, well dressed, waiting for someone for over 25 mins & visibly distressed. Spoke to her, she was on an interview and didn't show up (stupid CEO). We spoke more. Told her what @boundaryML does,…
End @github Universe ‘25 with a bang! 🎉 Join us for an evening of AI demos, startup stories, and great conversations with builders shaping the future of software. Hosted by Github for Startups, Flox, and @AI21Labs, this event brings together the community’s most inspiring…
I put the @boundaryML and JSON Schemas payloads side by side. Turns out that BAML is much lighter to serialize and parse. Aggregate (all tests): 84.4% reduction - 🔴 JSON Schema: 632 tokens - 🟢 BAML Schema: 99 tokens Use BAML in your Ruby apps with github.com/vicentereig/so…
Great event with @mlopscommunity at the @boundaryML offices - thanks to them for hosting and sponsoring the food/drink. Enjoyed Michael’s talk about Tambo and agentic UX, and Vaibhav’s talk about BAML language integrations - exciting to see how far both companies have come…
Stay tuned!
BAML hit 6,000 stars on Github today. 10k coming soon once we drop agent support in the next 2 months. Imagine iterating on agents / ai workflows 10x faster.
We are hiring Rust devs to work on building out a programming language (BAML) with crazy new, useful features for making AI pipelines. Our entire team is senior+ engineers. We are a team of 5. We work hard and in person in Seattle. We accept new grads. The interview is not…
The Rise of Semantic Entity Resolution is on Towards Data Science this morning…! > This post introduces the emerging field of semantic entity resolution for knowledge graphs, which uses language models to automate the most painful part of building knowledge graphs from text:…
DSPy 🤝 BAML
DSPy users can now leverage BAML's output format. It can give you 5% gain in structured output reliability, even in bigger models. from dspy.adapters.baml_adapter import BAMLAdapter Great work from our community member @tech_optimist . BAML users have had these reliability…
also we used @boundaryML to build this and it is awesome github.com/jazzberry-ai/s…
In case you were doubting BAML's reliability with structured outputs, a BAML user got SOTA on agentic web-browsing. Think of all the dev cycles spent trying to use tool-call APIs for 10 different providers, when they could've just used BAML.
Magnitude is state-of-the-art on WebVoyager at 94% 🚀 Here's a video of me using Magnitude to generate a report based on our WebVoyager blog. Here's what our agent said: { title: 'Magnitude WebVoyager Benchmark Results', summary: "Magnitude has achieved state-of-the-art…
Keep the BAML projects coming
When I worked at @TetraScience I learned how to use @elastic. Something I found a bit strange, however, is just how poorly documented regular human queries were paired to ElasticSearch (EQL) language queries. I decided to do something about this.
Home automation is a >$100B industry With open source LLMs on @CerebrasSystems, anyone can run blazingly fast inference basically for free so I remade the core of Alexa in a day.. using @boundaryML @elevenlabsio @convex_dev A thread 🧵 (with github repo!)
was living under the rock to have never heard about BAML. thanks to @tech_optimist for introducing me to BAML, have been digging into it for the past 2 days, lives up to the promises for structured extraction with complex json or pydantic schema, BAML is the way to go for sure.…
I built an open-source deep research agent with Gemini, @ExaAILabs and @boundaryML's BAML + a terminal UI (also has a separate @inngest backend which I will do another video on later) Check out how it works 👇
Its crazy how easy @boundaryML has made making AI agents. After trying Agno, langchain etc I am convinced BAML is the best way to build AI Agents
What if you did all LLM evals on a spreadsheet? Made with BAML by AndrewD on our discord
Read how we are changing the game in LLM Observability
If an API failed 5% of the time, engineers would fix it immediately. So why do we accept that from LLMs? @boundaryML is changing that with BAML, a typed programming language for LLM workflows—and @ClickHouseDB as the observability layer behind it all. clickhou.se/4kkY1TE
👩💻 Still writing LLM prompts with f-strings? That might’ve worked early on — but today’s AI apps demand structure, speed & observability. BAML (created by @BoundaryML) is here to change the game. 🧵👇
United States Trends
- 1. Megyn Kelly 23.3K posts
- 2. Vine 25.8K posts
- 3. Senator Fetterman 14.5K posts
- 4. CarPlay 3,859 posts
- 5. #NXXT_JPMorgan N/A
- 6. Osimhen 90.3K posts
- 7. Cynthia 106K posts
- 8. Padres 28.8K posts
- 9. Black Mirror 4,721 posts
- 10. Katie Couric 8,434 posts
- 11. #WorldKindnessDay 16K posts
- 12. Rep. Eric Swalwell 4,420 posts
- 13. Woody Johnson 1,419 posts
- 14. Gaetz 10.4K posts
- 15. Sheel N/A
- 16. Gabon 117K posts
- 17. V-fib N/A
- 18. Bonhoeffer 4,601 posts
- 19. VOTAR NO 26.5K posts
- 20. Cal Raleigh N/A
You might like
-
Cosine
@CosineAI -
Dishank Bansal
@theshank9 -
Apollo Research
@apolloaievals -
Founders Lounge Uganda
@FoundersLounge_ -
Rebekah Wanic
@BekahKnowsBS -
LiteLLM (YC W23)
@LiteLLM -
Truewind
@truewindai -
Rubbrband
@rubbrbandHQ -
Art of VC Fundraising
@fundraisetips -
Jorge Hernandez 🇺🇦 🏳️🌈
@braneloop -
Dustin
@dustintech -
Openlayer
@openlayerco -
Vasu Raman
@likeavass -
Ron I רון
@Ch00by -
Ishan Gupta
@ishangpta
Something went wrong.
Something went wrong.