The best thing about coding is having the power to create anything you can imagine. It’s just the bugs that get in the way.
Fixing a bug is like being both the detective and the killer in a mystery. After hours of searching for clues as to why the model crashed, you find the murder weapon: a RuntimeError: size mismatch at layer 42. And the motive? You were just being careless with your tensor shapes at 3 AM. Moments like these make me think living in caves wasn't so bad. 😤
But the victory of finally crushing those bugs—and seeing the loss curve drop—is the sweetest of all. 😎
When I'm not playing detective, here is what I actually do:
- Background: Final-year Statistics and Data Science undergrad (ML Minor) at IIT Kanpur.
- AI/ML Focus: LLMs, Transformers, Reinforcement Learning (RLHF/DPO), and evaluating Knowledge Graphs.
- Systems & Infra: CUDA/GPU optimization, multi-GPU training pipelines, and DBMS internals (B+ trees, join optimizations).
- Experience: Former Developer Intern at Samsung Research Institute Bangalore.