Software engineer at Amazon's Finance Automation org in Seattle. Working on using AI to automate various areas inside Amazon's finance division. Feel free to look around — you'll find projects I've worked on for fun, past research, and more.
A real-time game advisor that runs entirely on local hardware — screenshot capture, Claude vision extraction, hybrid retrieval over a verified knowledge base, and persistent game memory.
Building a decoder-only transformer from scratch to generate Shakespeare-style text. A deep dive into multi-head attention, positional encoding, and the magic of "Attention Is All You Need".
Deep dive into creating a Byte-Pair Encoding tokenizer from scratch, training it on Jayson Tatum Wikipedia data, and building a real-time tokenization interface.
Building a custom MLP architecture with Batch Normalization for character-level text generation. A deep dive into neural network fundamentals and the magic of BatchNorm.
Dive deep into the core mechanics of neural networks by building a custom automatic differentiation (autograd) engine and a Multi-Layer Perceptron (MLP) from the ground up.
Welcome to my personal corner of the internet! Explore my projects, get to know me, and see what I've been working on.
Leading the development of conversational AI agents that handle inbound calls for car dealerships, from proof of concept to one of the startup''s main selling points.
I built a decentralized auction platform using zero-knowledge proofs at Nokia Bell Labs, where you can prove you won without anyone learning what you bid.
Using GANs to synthesize particle accelerator data at Lawrence Livermore National Laboratory, because running an accelerator is expensive and physics needs more data than reality can provide.