Janit Anjaria

University of Maryland in the top 3 in the US. 🔥

Historic. Nostalgic. Epic! Lets go!!!!!

We put in the early mornings, late nights, and hard work to get to this far – but this is just the starting line. Self-driving freight is on the way.

Aurora and Volvo are not just gearing up to deploy driverless trucks; we are already delivering value to customers today. Our co-founder and Chief Product Officer @sterling_a Anderson joined Volvo Group CTO Lars Stenqvist on stage to showcase how our partnership is redefining leadership and transforming long-haul logistics with greater efficiency and safety via autonomous trucks. Watch their full segment:

Three leaders, one vision. Aurora, Continental, and NVIDIA unite to unlock a future of scalable autonomous freight. Read more here: bit.ly/3Pn6BUr

This is awesome.

I'll be recruiting PhD students this cycle! If you're interested in working on program synthesis, type systems, program analysis, and in general ways to make programming better, please reach out. Here some things we are working on: ku-progsys.github.io RTs appreciated 🙏

Strong picture!

@AkashTandon @OpenComputePrj IIRC this project has helped in a lot of protocol definitions across the network stack, as well as design choices. I think it’s a hit, but maybe not as hyped like the new age AI tooling?

TIL about the fascinating @OpenComputePrj - an initiative to share designs + best practices of data center hardware and promote sustainability. en.wikipedia.org/wiki/Open_Comp… Started as a Meta project - another one of their OSS hits? (unsure about its significance as of now) 👀

Curious to learn if we know how Google uses GShard (arxiv.org/abs/2006.16668) if at all it does? Considering how it seemed quite ahead in its approach to use MoEs?

@llama_index The paper is very interesting. The concept of closed-book and open-book is quite intuitive. Thank you @ravithejads for the implementation in LlamaPack.

RAFT - Retrieval Augmented Fine Tuning 🔥 RAFT offers a method to fine-tune pre-trained LLMs for specific domain RAG settings. Conventional RAG is like an open-book exam, retrieving documents from an index to provide context for answering queries. This makes it more effective than the closed-book exam setting where LLMs rely solely on their pre-training and fine-tuning to respond to prompts, but doesn't allow the LLM to learn the domain beforehand. RAFT from @tianjun_zhang, @shishirpatil_ proposed a fine-tuning approach that enhances LLMs for domain-specific open-book exams, where the model is trained to attend to relevant docs and ignore irrelevant documents. RAFT creates a synthetic dataset where each data sample consists of: 1️⃣ A question 2️⃣ Two docs: An "oracle" document relevant to the question, and a "distractor" documents irrelevant to the question. 3️⃣ An answer generated from the documents 4️⃣ A Chain-of-Thought explanation including excerpts from the relevant documents. The created synthetic dataset is further used for fine-tuning to improve RAG performance. Thanks to @ravithejads, you can now generate RAFT dataset using our brand new RAFTDatasetPack LlamaPack. LlamaPack: github.com/run-llama/llam… Video Tutorial: youtube.com/watch?v=sqPckk… Paper: arxiv.org/abs/2403.10131

I desperately miss good old days of ML, when discussions were more technical and measured. The papers and releases were full of ideas and details rather than fancy gifs/videos. Can the "money guys" find something else to keep themselves busy?

I think this might be a knee-jerk reaction to a miscommunicated process. Startups might not really have much impact, but it might mostly be around BigTech. Source: indianexpress.com/article/busine…

Not sure if this is a good move. Maybe the intent is good, but this sounds like detrimental to the ecosystem as a whole. A regular model with 7Bish params needs a government approval? 🤯 is there any other explanation?

@Rahul_J_Mathur Would highly recommend @foldmoney_ from @meakaakka and gang.

A major problem which PFM apps in India face is that they misread SMS notifications For eg - for each > ₹2K txn on my CC, Kotak Bank sends me at least 2 notifications asking me to convert it to an EMI And, the PFM apps misread this as debit transactions (i.e. they triple count the same spend) Getting CC transaction level data on AA (Account Aggregator) rails may solve for this beautifully.. But, I think the jigsaw puzzle is coming together - as more data is available on AA, the holy grail of PFM may just get cracked in India - lots of interesting cos solving for this.

I have been working with this team and the product for quite sometime now. One thing that the team does best is the love and precision to details in their work. Now that it is out of Beta, I would say go to your App Store and download the app NOW. @meakaakka

At IGI Airport New Delhi- it is insane how spending power has increased. I remember how eating or buying anything was a status statement growing up. And now cafeterias and restaurants are packed. Great to see increase in earning potential. 🇮🇳

Auto-Tagging is one of the fundamental features of @foldmoney_. We spent much time making this auto-tagging layer intelligent and scalable. Some tech notes on how we built this (TLDR: Embeddings + vector database).

1975 is remembered for Sholay. But another Dharmendra-Amitabh Bachchan starrer released that year and is considered to be one of the finest comedy films of all time. A thread on Chupke Chupke. 🧵

Though this looks like an awesome end product, the real advancement is in terms of the individual pieces of technology that empower this kind of an end product. #technology #Robotics #AI #teslaBot