I am Sachin. I work on Deep learning and Reinforcement learning. I think AGI will be DL + RL fused together.

I want to innovate frontier intelligence and "research which scales". innovation where i need to think about newer methods of architecture and model training and also to scale that acorss lot of gpu's.

things which i am doing and interests me.

• scalabe distributive training of DL
• long horizon RL for language models.
• DL/RL infra

what i find fascinating currently:

• anything on intelligence and learning
• realizing Brahman and self

at broader computation fascinates me, i feel like computation is foundational property of universe. it could be that physics is informational. this gives me realization that any state of matter treated as information can be compressed and manipulated through intact computation algorithm.

I remember words of Max plank in his autobiography that "laws of human reasoning coincide with the laws governing the world...therefore, pure reasoning can enable man to gain an insight into the mechanism of the latter." Our mind can reach into the fabric of universe. I have curiosity to find this laws for intelligence and learning, for me i think its single most important quest.

core principles and ideas.

• “It just happens they got interested in this thing, and they learned all this stuff. They're just people. …So if you take an ordinary person who's willing to devote a great deal of time and study and work and thinking and mathematics, then he's become a scientist.” I think about this Richard feynman lines a lot.
• everything that is "understood" can be "understood"
• be unreasonable yet truthful.
• Choose reality over consensus and language. ask for law of nature over social eqillibrium.
• reality is projection which appears in self(consciousness); we never have equation for converting mass to consciousness.That bramhan is from which all appears.

I enjoy reading books, chess(learning with closed eyes), playing piano, creating things with ambitious fellas, rabbit holes, meditation, solitude, stargazing, winters, lush greenery, aravalli's mountain range, bitter lesson, solving programming problems, o3, discussion with gpt pro models.

i admire: Gunvant shah, Richard feynman, Yajnavalkya.

Courage and curiosity

i think courage to ask bold question is important. seeking truth and creation requires curiosity with courage. courage here is choosing nature over consensus, not embracing something untill it right and inevitable.

naturally i fall in love with science and engineering, i cultivated it throughout my childhood. i had small laboratory where i collected pieces from scrapes and broken toys. i use it to make models of rockets, spaceships, drones and even satelites its spiritual thing for me when out of all this scrapes wonderful creation can be done from imagination.

i want to understand this fabric of universe and put that into creation just like i did that with scrapes and motors. imagination understanding and innovation.