Projects

Real-time visualization of LLM internals — trace token generation, attention patterns, hidden states, and probability distributions as they happen. Built to understand what's actually going on inside these models.

Experimenting with Coinbase's x402 protocol — HTTP-native micropayments on Base using USDC. Server middleware intercepts requests, client auto-signs payments. Exploring what autonomous agent-to-agent commerce looks like.

Web-based platform for training and fine-tuning LLMs. Experiment tracking, dataset management, real-time loss curves, and an interactive playground to test checkpoints mid-training.

Document intelligence pipeline for digitizing Indian manuscripts and heritage texts. Integrated Sarvam AI's document API for OCR across 22 Indian languages + English.

Data extraction tool for archiving Discord channel history into a local SQLite database. Built for preserving community knowledge.

Running and benchmarking local LLMs with llama.cpp. Measuring inference speed, memory footprint, and quality across different quantization levels.