
Archie Sengupta
@archiexzzz • 67,642 subscribers
ai researcher (robotics & world model) // pre & post-training // prev: @atomicworkhq, 5x yc dev
Shorts
Videos

Are you kidding me? This is the best website design experience I've had in a long, long time. Shopify just won my heart with this new Winter Edition web design! Seriously, the bar for good design is so high right now. With all the AI slop out there, the only word-of-mouth you can rely on is having great design taste and building a cult around it. Just take Prime Intellect - ultimate design taste. Why wouldn't I send this to my friends or post it online organically? The moat is good design, folks - it's so important with all the AI slop we're seeing today. tobi lutke please give your design team a big big raise!
Archie Sengupta239,145 次观看 • 5 个月前

Every now and then, I come back to this interview of Shah Rukh Khan on how to be successful(for the younger generation): "you can't be restful and at peace if you want to be successful and it's the truth, it's god's honest truth." "jo bhi apko bole aram karo, zindi jo h is beyond work and all, its good that you donot want to be successful. i feel envy and envious of people who relax and rests. i feel jealous of people who dont want to be successful". learn from the greatest. always.
Archie Sengupta65,840 次观看 • 1 年前

vLLM fast inference running on monte carlo synthetic data generation with: > peak generation throughput of ~ 23k token/s & avg of 20k token/s > ~200 reqs/s > Qwen/Qwen2.5-0.5B-Instruct > on 1x Grace Hopper 200 - 480 GB(~96GB HBM3) vllm config: --max-num-seqs 512 --chunked-prefill-enabled (for better throughput) --dtype float16: (half precision for mem efficiency) --gpu-memory-utilization 0.95(for better KV Caching)
Archie Sengupta26,130 次观看 • 5 个月前
没有更多内容可加载

