Background

Third-year CS student based in India. Competitive programming got me into this. Expert on Codeforces, Knight on LeetCode, 4-star on CodeChef. That habit of thinking a problem through before writing any code has stuck, and I apply it to everything I build.

I contribute to open source seriously. Google Summer of Code 2026 developer on the CGAL project since May 2026, working on OpenGL shader systems in C++. I've also contributed to Accomplish AI, OpenMetadata, KubeEdge, and others.

What I Build

Web side: React, Next.js, Node.js or Python backends, PostgreSQL. I pay attention to the details most people skip. Motion, spacing, interaction states. Whether something feels deliberate or just default.

AI side: I go deep on LLMs. I've built GPT-2 (124M params), Llama 3.2, Qwen3 MoE, and Gemma 3 from scratch in PyTorch. I work on KV cache compression, fast inference with Triton kernels, and agentic systems. The goal is always something that holds up under real use, not just a demo.