Hi 👋, I'm Krishna! and welcome to my dev space.
I build machine learning algorithms, compiler infrastructure, and high-performance systems from the ground up. My engineering philosophy is simple: understand the stack from the hardware execution to the mathematical primitives. No bloated abstractions. No black boxes. Just optimal Time/Space complexity, rigorous system design, and clean C++/Python.
- ArchiMind: A multi-agent system utilizing LangGraph and ChromaDB to parse complex codebases and autonomously generate chapter-wise architecture documentation, HLD/LLD, and Mermaid diagrams.
- Graphite: A full-stack web UI design and preview tool featuring a React/ShadCN frontend and a FastAPI backend with dynamic component rendering capabilities.
- Arashi Browser: A custom, high-performance web browser engineered for novel navigation paradigms and optimized rendering (Prototyped during a Google Hackathon).
- zero-abstraction-build-it-all: Implementing core CS paradigms and ML primitives entirely from scratch in C, C++, and Rust. Zero library imports. Pure logic.
- AriaLM: A ~5M parameter LLaMA-style Transformer trained from scratch. Implements RoPE, Grouped Query Attention (GQA), SwiGLU, and RMSNorm for symbolic music generation.
- chess_reinforce: AlphaZero-inspired Chess AI built with a custom Transformer encoder and Policy Gradient self-play.
- linux_sleep_driver: System-level Bash utility for advanced suspend-then-hibernate power management directly interfacing with ACPID.
- Languages: C, C++, Python, Rust
- ML & Compute: PyTorch, TensorFlow, OpenCV, Scikit-Learn (Open Source Contributor)
- Architecture: Hardware-Software Optimization, SIMD Abstractions, Multi-Agent Orchestration, Kernel/Driver level configurations
