gautham (@capemox) header
gautham (@capemox) profile picture

angle born in a convex hull

Bengaluru, India
Joined July 2023
Photo by gautham (@capemox)Photo by gautham (@capemox)Photo by gautham (@capemox)Photo by gautham (@capemox)Photo by gautham (@capemox)Photo by gautham (@capemox)Photo by gautham (@capemox)Photo by gautham (@capemox)Photo by gautham (@capemox)Photo by gautham (@capemox)
At “Women in Systems”, we will feature a panel discussion on the evolution and broader development of systems with Sujatha Sivaramakrishnan, Meera Sundar, and Janhavi Tripurwar at Microsoft. Registrations close on Tuesday, the 24th. See you there! hasgeek.com/bengalurusystems…
7
20
893
We are thrilled to host a special meetup, "Women in Systems", at Microsoft (Bellandur) for our November edition on the 28th, featuring lightning talks as well as a panel discussion. This meetup is open to everyone, so feel free to register! hasgeek.com/bengalurusystems…
11
2
43
2,372
That's a wrap for the October Edition, this time with @lossfunk! Here's what went down: - @kaxsthxbh walked us through how inference works from first principles. - @DivyanshGolyan gave us a look inside Rui, a Zig agent harness. (1/2)
1
7
1
49
2,123
Just sparsify everything! Insane release by Paulo and the blog post is a delightful read, give it a go!
We surfed the dense and late-interaction waves… now it's time to SlideOn sparse retrieval 🐕 🚀 Meet SlideOn and mSlideOn, fully open SOTA sparse retrievers at 174M and 339M parameters, trained by @LightOnIO R&D Models: huggingface.co/collections/l… Blog: huggingface.co/blog/lightona…
1
7
21
814
I'm talking about Inference Free SPLADE models at @softwaredoug's Retrieval Augmented Gathering today. SPLADE uses a transformer neural network to expand your document's keywords, so keyword search can match synonyms (and other semantically similar keywords). But it runs a model on every query, which adds upto 50ms p50 latency on CPU and needs a GPU to be faster. Inference-free SPLADE skips that step. It does more model work upfront on your documents, so queries cost about as much as BM25. We'll talk about how it works, what you trade off, and how to take it to production :) Join today if you wanna learn more! 📺👇
2
6
19
857
Introducing SPARSEUP: the first model release from @Linkup_platform , and the missing sparse companion of DenseOn and LateOn. Same backbone, same data, <150M, 56+ on BEIR-13. The model is open-source, use it! Blog: linkup.so/blog/introducing-s… Model: huggingface.co/Linkup-Platfo… (Apache 2.0)
11
21
13
118
87,633
"Multilingual when?" Glad to finally answer today! Introducing mDenseOn & mLateOn, the next generation of fully open models for multilingual, long-context, and code retrieval Same openness, same frontier performance, but now for everything
12
37
7
115
14,107
Today, we release a collection of multimodal rerankers, handling both text & document images powered by @coreprinciple_ 's insane work! The 2B kinda mogs other same-size open rerankers 👀 (yes, there are so few of them 😭) 62.66 NDCG@10 on ViDoRe V3! Here’s what we learned 🧵
🤖 Made with AI
3
14
4
49
9,605
Very interesting read Search is changing these days, the trade-off we are willing to make are different Something that was considered way too long processing in the past may be saving you a lot of time nowadays It does not mean you should spend hours per search call, it just means that the quality of the search has even more weight than before and sacrifying _a bit_ of latency to save 10 turns is a net win I would also go one step further: what's slow today might not be slow tomorrow. We, as a community, are pretty good to optimize things (even more so with LLMs!) We should explore and find the most promising trade-offs, the infra will follow
9
79
6,838
gautham retweeted
Replying to @capemox
Exactly this. I really don’t think the availability of infra should have any bearing whatsoever on whether an idea makes it into the world. We need to encourage papers whose infra is in the future
1
2
9
727
gautham retweeted
Even strong multi-vector models may break down when optimized for low-latency and high-QPS inference in production. But this can be fixed. We're open-sourcing Iso-ModernColBERT, a late interaction model built for efficient inference and scalable retrieval. 🧵 (1/6)
1
9
7
55
10,774
gautham retweeted
🤗 Announcing the Ettin Reranker family: six new CrossEncoder rerankers from 17M to 1B parameters, state-of-the-art at their respective sizes. Built on the Ettin ModernBERT encoders, with the full training recipe and ~143M-triple training dataset as well. 🧵
10
29
6
142
26,007
With just a few days to go, we have a limited number of seats available for our panel discussion on ‘Correctness in AI-Accelerated Systems Engineering’ /w @AntithesisHQ! Register here! hasgeek.com/bengalurusystems…
1
6
38
2,088
gautham retweeted
🤖 I've just released Sentence Transformers v5.5.0! It's headlined by a new `train-sentence-transformers` Agent Skill: let your AI coding agent train & finetune embedding, reranker & sparse encoder models for you. Plus more, here's a thread with highlights 🧵
3
14
1
79
3,726
We're featuring a panel discussion on ‘Correctness in AI-Accelerated Systems Engineering’ with Will Wilson, the CEO of @AntithesisHQ and @amodm, co-founder of @pre6ai, moderated by @AnirudhRowjee from @couchbase on the 22nd! Register to attend! hasgeek.com/bengalurusystems…
6
1
42
2,389
gautham retweeted
One retriever for real world data. SOTA on BEIR, built for production. LightOn open-sources LateOn, its multi-vector retriever. Apache 2.0. 149M parameters. LateOn powers LightOn's own search pipeline. Built and deployed in the best hybrid search stack on the market. Why is LateOn the retriever you actually need? We decontaminated BEIR, to remove training overlap, keep only the unseen. Late interaction compounds: LateOn tops that leaderboard too. This is the closest eval to your private data. 🔗 Read the full story: lighton.ai/lighton-blogs/the… 🔑 And claim your API key to test it now: lighton.ai/get-access 👨‍🍳 Kudos to @raphaelsrty, @antoine_chaffin, @orionweller, @paulomouraj, @AmelieTabatta !
2
16
1
58
6,187
Bengaluru Systems is back with the meet-up on 25 April, this time @Zerodha! Our speakers include Mithun Chicklore Yogendra from @ParadeDB and Sourav Singh Rawat from @Cloudflare! Register to attend - hasgeek.com/bengalurusystems… See you there!
2
7
2
72
6,517
And that's a wrap for March's Edition of Bengaluru Systems! Here are the highlights: - @AnirudhRowjee talked about LevelDB's write path and a high-level overview of its architecture. - @capemox discussed how Bleve's full text and Vector Search indexes work. (1/2)
1
3
26
1,656