Kanishka Das

Qualified GATE 2026 > scored 37.99 with 12.5k rank in CSIT > scored 26.67 ( haha 67 ) with 9.5k rank in DA > was a trial run > paper was easy ( i think so i barely did any pyqs so i cant compare ) > next year i'm confident imma get something good ( ts easy ) > hid my picture in the scorecard cause i look so chopped if you still have some job for me that has good money please reach out. ( i need money).

won innoviz 😹🤙 only thing that would make me happier rn is one of those very high paying internships 😞💔 (Please reach out if you got too much money) In hindsight, had a really good time though. Two sleepless nights totally worth it, what was more impressive than the actual implementation was the improvisation and optimising the software due to hardware limitations. Will probably make a detailed technical video going over the software implementation. Esp32 and hardware done by Ravnish Kumar and website built by Kashish Raheja , software by yours truly. Also thanks to St. Andrews Institute of Technology & Management for hosting this🤙🤙

built something that genuinely surprised me when it worked. Project Niyom — it discovers physical laws from simulation data  using sparse regression. the idea: simulate a physical system, collect trajectory data,  then pretend you don't know the equations. can the machine  figure them out from just the numbers? turns out yes. for the Simple Harmonic Oscillator it recovered F = -kx  with 0.19% coefficient error. for the Lorenz attractor —  a chaotic system with three coupled nonlinear equations —  it recovered all three with under 3% error. the interesting engineering problem wasn't the math.  it was building something that works on ANY system. so i designed a file format called .niy where you write  your equations in plain text: no code, just physics notation.  the pipeline reads it, simulates it, and runs the full  SINDy discovery automatically. drop lorenz_attractor.niy in. get the butterfly equations back. one thing i learned that's worth sharing: Mean Absolute  Error is the wrong validation metric for chaotic systems.  even if you recover the equations perfectly, two Lorenz  trajectories diverge exponentially because of the butterfly  effect. the right check is whether the discovered system  traces the same attractor shape — and it does. built as my BTech minor project. considering writing it up  as a paper. the .niy format as a standard DSL for dynamical  systems feels like the part worth publishing. try it out here: https://lnkd.in/gACfFksX