Final-year MCA @ NIT Warangal
Building evaluation pipelines for LLMs · Writing C++ that has to be fast · Playing basketball otherwise 🏀
Chess Analysis and Learning Tool (CALT) · MIT License
A research pipeline that asks a hard question: when an LLM explains a chess position, is it actually right — and is "right" even the same thing as "instructive"?
- Benchmarks LLM-generated commentary against Stockfish ground truth and blind human annotation across the 300-position Win-at-Chess set, scored on a four-axis compound metric
- Retrieval subsystem pairing bitboard tactical-motif detection with Zobrist-anchored semantic lookup, so commentary is grounded in verified patterns instead of free generation
- FastAPI + SSE streaming over a multi-process engine pool; Gemma and Qwen served locally via
llama.cppagainst a shared cache to isolate model quality from engine variance
Python · PyTorch · FastAPI · llama.cpp · Stockfish · SSE
Error-Correcting Codes · C++17
Link-layer error detection and correction, from scratch, zero dependencies. Hamming codes with syndrome-based single-bit correction and CRC via polynomial division — each scheme validated against injected bit-flip patterns.
C++17 · Computer Networks · Bit manipulation
1,300 in C++ · 500+ day streak · top 1% of rated contestants.
Six published technical articles for LinuxForDevices on the Rust toolchain, Pyright static type checking, and Linux desktop tooling — written for developers who want the why, not just the command.
Final year of my MCA at NIT Warangal, graduating 2027. Deep in LLM evaluation, systems programming, and competitive programming.
Open to full-time SDE, ML engineering, and AI research roles for the 2027 batch. If you're hiring or just want to argue about whether Stockfish's best move is ever the right move to teach — reach out.