arun @ nyu

slowing down //
cs @ nyu //

[arun @ nyu @ArunPatro]

@nishnik @Meta Amazing! Congrats Nishant!

[rahul @rahulgs]

we're hiring full stack engineers to work on llms at ramp

if interested, dm me with examples of real things you've built

come work on real deployments and learn how to drive enterprise value

we're a small and mighty team

what we've worked on in the last year:
- multi-step agents for document extraction / ocr (sota accuracy, probably). llm agents + constraint solvers
- low-latency next action prediction in our web app (more soon)
- ramp tour guide: https://t.co/p6dRYIuZFl
- web agents for solving c*ptchas
- codebase import cycle removal with ast parsing/graph cutting algorithms + llms in our python monolith backend
- sales outbound automation and lead scoring agents
- llm model routing between third party providers (+per feature cost tracking)
- llm infra: embedding/reranking/generation finetuning and on-prem deployment/inference
- structured extraction (https://t.co/aDz5Dc9RTY)
- customer feedback extraction from meeting recordings / routing to marketing
- internal tools for: underwriting team/product team/sales/customer support teams
- global search / function calling copilot
- receipt matching (retrieval)
- sms llm interface (function calling)
- suggested memos
- automated accounting coding
- natural language report generation
+ more

[rahul @rahulgs]

Shoutout to Ramp engineer Andrew Gu who was on the coaching staff.

Congratulations to Team USA for winning IMO 2024!

[hu-po @hupobuboo]

thermodynamic computers bring the heat https://t.co/cEd0TXjbPa

[arun @ nyu @ArunPatro]

I have been waiting long to finally try this out! BayesLLMs go brrrrrr. Kudos to @Sam_Duffield and co

[Normal Computing 🧠🌡️ @NormalComputing]

We're excited to announce posteriors!

posteriors is an open-source Python library designed to make it as easy as possible to apply uncertainty quantification to deep learning models with PyTorch.

https://t.co/8fVhaTnr5b

[arun @ nyu @ArunPatro]

A huge thanks to @RamiSbahi @dmitri_saberi @thomasahle for helping drive this! We are also extremely grateful to @paradigm for Flux, upon which we have built Branches.

[arun @ nyu @ArunPatro]

Search methods like Tree-of-Thoughts have become immensely popular for solving problems that often fail with direct prompting techniques like Chain-of-Thought. In our internal applications of graph-based search methods, we found these methods difficult to debug and interpret.

[arun @ nyu @ArunPatro]

To learn more about how we tried to make tree methods faster, more accurate, and more collaborative, take a look at the full writeup!
https://t.co/fIQe0Ewk5c

[Phoebe Zeng @pl_zeng]

This week's installment of SLT in the park, keeping warm w/ Watanabe's free energy. @MaxAifer @dmitri_saberi @ArunPatro @NormalComputing

[Normal Computing 🧠🌡️ @NormalComputing]

Extended Mind Transformers (EMTs) are a new approach to working with very large contexts and external data sources developed by @KlettPhoebe, @thomasahle, Normal's AI team. Inspired by the Extended Mind Thesis, we modify Multihead Attention to directly query a vector database.

[arun @ nyu @ArunPatro]

Thursdays are for Singular Learning Theory by the park!

[arun @ nyu @ArunPatro]

@dmitri_saberi @dan_p_simpson @dberri18 😬

[Normal Computing 🧠🌡️ @NormalComputing]

At Normal, we build full-stack objective-driven AI systems capable of reasoning in the real world.

In our latest Blog, we examine explainability techniques for language models that we’ve found useful for improving reliability in reasoning.

👇👇

https://t.co/kknhfWH5v7

[arun @ nyu @ArunPatro]

Explainability of GenAI is a critical step towards reliable and informed decision making. Read more about Normal's endeavors to make GenAI more trustworthy as we solve problems on hallucinations, reasoning and referencing. #xai

https://t.co/4jGvILKrMN