sanchay thalnerkar

sanchay thalnerkar

i build agents, i automate work, and i build and train models.

  1. dotminds · first real codebase. frontend intern. react and react native.

  2. acmegrade · first ml internship. data cleaning, classical models, a lot of pandas.

  3. creative finserve · django apis and a crm for a finance firm.

  4. creative finserve · data science intern.

  5. ai intern

    then i learned in other peoples rooms.

    ai intern at intel.

  6. computer vision and ai interndec 2022 to feb 2023, remotedriver assistance: lane departure, adaptive cruise, object tracking.on linkedin

    computer vision at ibm.

    teaching a model where the cars around you are: lane departure warning, adaptive cruise, object tracking.

  7. unscript 2023

    then one of them counted.

    first in ai/ml, for a model that scores how easy a document is to read. the judges did not know me.

  8. 6+ hackathons, most on no sleep. a writing coach, an audiobook maker, a smart home agent, a parking spot finder.

  9. lablab.ai · tutorials for people learning what i was. llama 3, groq and gpt-4o builds, written up.

  10. two papers published.

  11. x · one thread, read by 1.8m people. what a reasoning model does with a chess board.

  12. then the work got real.

    bank statement engines and fraud checks for a fintech in india. a wrong answer costs someone money, so i measure before i trust.

  13. now, a fleet of agents.

    they write most of the code. i decide what is worth writing, and check what comes back.

theusefulthingwasneverthemodel.itwastheworkthemodeltookoffsomeonesdesk.

so i turned to agents and automation. first i automated my own work, then other peoples. now i run fleets of coding agents every day. they write most of the code, i decide what is worth writing and check what comes back.

right now i build ai booking and automation systems for clinics and businesses, from a hearing clinic group in australia to a fintech in india. one person, working like a team.

three things i build.

  1. agentsi run fleets of coding agents. they write the code, i decide what is worth writing.
  2. modelsi build and train models where a wrong answer costs money. fraud checks, documents, vision.
  3. automationi automate the work businesses still do by hand. clinic bookings, bank statements, follow ups.

one person, working like a team.

built in mumbai, for teams in san francisco, dubai and sydney

right now

building ai booking systems for hearing clinics in australia, with a fleet of coding agents doing most of the typing.

every build, and how to check itevery build, and how to check it

blogs

  1. my agents write the code. here is what i still do myselfoct 2026a fleet of coding agents does most of the typing now. the deciding, the checking and the saying no are still mine.
  2. the boring work is where the money issep 2026clinic bookings and bank statements taught me more about ai than any demo. nobody brags about them, everybody pays for them.
  3. the o1 chess thread, and what it showedjan 20251.8m people read it. what a reasoning model does when you hand it a board and get out of the way.
all blogsall blogs