datavorous

Lightweight Vector Search Library focused on Memory Efficiency

a ~1200 Elo chess engine that fits within 2KB

A model specialized CPU inference runtime for Sarvam-1 written in C

A multithreaded, BVH enabled, cpu ray/path tracer

Various coding challenges from all around the internet

documenting my work in inference engineering

Yet Another Reddit Scrapper (without API keys) | Scrap search results, posts and images from subreddits filtered by hot, new etc and bulk download any user's data.

Distraction blocker which monitors your mouse and keyboard. Prevents you from switching tabs and wasting your time.

neofetch + falling sand engine for your terminal

A blazing fast, low footprint pixel art editor built with raylib, featuring a customizable canvas, multiple color palettes, and frame management for animations.

A tiny path tracer written in vanilla python to produce realistic renders

a micro inference runtime to run llama 2 models

a tiny chess engine playground

Reaction-Diffusion simulation using the Gray-Scott model, available in both C and web versions. It visualizes various patterns that emerge from the interaction of two virtual chemicals in a 2D grid.

the long awaited physics engine of datavorous

TFLite graph analyzer for QNN delegation diagnostics.

a toy vector search library in python

frogemm is a small GEMM library to do fast matrix multiplications

A collection of prompts for responsible AI usage

faithful re implementation of GPT-2

A lightweight boids simulation showcasing emergent flocking behavior

token factory simulation to do inference optimizations

karpathy's micrograd but had to give it a different name

search PDK

A clean minimal boilerplate to setup a test bed for experiments.

implementing kvcache from scratch (benchmarked against gpt-medium)

a dead simple linux syscall tracer in C

No description.

ai voice agent [hackathon project]

hackathon project :: MCP to render bar/pie/line/scatter plots, find leads from reddit, gets med info from 1mg, and provides persistent memory to store TODOs

Accessible large language models via k-bit quantization for PyTorch.

No description.

a quantization library

sequence statistics and k-mer analysis tool for FASTA files (WIP)

experimenting with cuda and learning something along the way

a voxel engine to cure my year-end boredom

lightweight garbage collector from scratch

a toy browser made from first principles