I wanted to reintroduce myself here in this platform:
i’m a 22 year old DS working at a HiLabs & contributing to GNU Octave through GSOC ’26
over the last few years, i’ve somehow ended up doing a bit of everything i genuinely enjoy:
1. 25+ merged PRs in Octave, mostly around performance & memory bottlenecks in systems like KD-Trees, kNN, HNSW etc
2. currently working on a research idea around hierarchical RAG systems & query-conditioned retrieval
3. taking part in the ICML AdaptFM challenge to make Qwen inference faster
4. built & experimented with RAG pipelines, multimodal systems, RL, LLM eval frameworks, fine-tuning & retrieval systems during internships/projects
5. won a few hackathons along the way & still love competitions because they force you to learn fast
6. ex-Codeforces Expert, led IIT Roorkee to a Silver Medal at Inter IIT Tech (ISRO Geospatial PS), still trying to find time for CP again
7. outside tech, i spend a lot of time running, lifting or making random edits
still learning everyday, but excited for what’s ahead. I hope I am reaching the right people who would help me drift in the right direction.
Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API.
Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls.
Try it: https://t.co/hhO6qTawgb 🐡
working with a fast pace startup gets you a lot of initial learning which would take you years to learn in a big MNC. Really grateful to work on problems which are tangent to my current skillset, around inference optimisations, llm hosting & gpu workload scaling in kub and much more!
BREAKING: GLM-5.2 is now 1st on Design Arena.
With an Elo of 1360, GLM-5.2 has jumped ahead of the now unavailable Claude Fable 5.
And it's open weights.
This is an improvement of 4 positions and 27 Elo points to achieve one of the highest Elo scores in our code categories since Design Arena started.
Huge congratulations to the @Zai_org on the release!
For well executed reasoning RL I would say:
https://t.co/0GUjcd1i3O
https://t.co/bbPlAy6iBE
https://t.co/FaxHUgkePT
https://t.co/TfIUlyhSQS
https://t.co/vwHoq4gKT0
https://t.co/2H9zqrHFpu
https://t.co/j4oiCXBGJn
https://t.co/D8Q2Qf0HqY
https://t.co/vdNO63mpDJ
https://t.co/uQZoQSKIkf
https://t.co/aIXppIO3xu
For more good stuff on infra (the others before also had good stuff):
https://t.co/pZzP35tyWQ
https://t.co/0uHfqcu424
For more agentic related:
https://t.co/aAilnJ0CRK
https://t.co/M5MKOrP3lg
https://t.co/9PGZ4Ku3hu
https://t.co/Iv82C6MRQq
https://t.co/qmgTSVI3C2
https://t.co/Wwl1cWZ6I8
For general RL and some cool ideas:
https://t.co/nyUuqHZ8zx
https://t.co/zirVXV3lNm
https://t.co/aoq1vhuBvQ
https://t.co/hyXOCRoqF4
https://t.co/H0lkrZLa8J
https://t.co/KDO7fqyLs9