
QuasiQ
A simple quantum computer simulator. Implements quantum gates, superposition, and entanglement.
Learn more →Things I've built.

A simple quantum computer simulator. Implements quantum gates, superposition, and entanglement.
Learn more →
A minimal autograd engine built from scratch in under 1000 lines. Capable of training complex models like GPT.
Learn more →
A minimal implementation of JAX-style function transformation engine. Builds automatic differentiation, JVP, and VJP from first principles.
Learn more →
The first Marathi programming language. A high-level, dynamically typed, interpreted language with syntax and documentation entirely in Marathi.
Learn more →
A minimal container runtime built from scratch, similar to Docker. Implements core container functionality using chroot and syscalls for process isolation.
Learn more →
A lightweight implementation of Group Relative Policy Optimization (GRPO), a reinforcement learning algorithm.
Learn more →
Working implementation of DeepSeek's Multi-Head Latent Attention mechanism.
Learn more →
A collection of autoregressive model implementations. Includes diverse models and experiments designed for easy replication and modification.
Learn more →An alternative approach to calculating self-attention using graph-based methods.
Learn more →A minimal library for Reinforcement Learning on Large Language Models. Provides clean abstractions for RLHF and related techniques.
Learn more →Using Reinforcement Learning to discover steering vectors for controlled text generation.
Learn more →