Senior at ASU. I spend most of my time thinking about distributed systems, reinforcement learning, and how to build things that don't fall apart at scale.
Most recently worked on data pipelines at Chunkr. Before that, built internal tools at Oats Overnight and automated data workflows at PrivateBlok.
This site is a playground. Right now I'm working on training language models to negotiate.
I built Bluecast, a SaaS tool for growing on LinkedIn before I walked away when the market got crowded. More recently, I've been training language models to negotiate through self-play RL and GRPO, fine-tuning with LoRA first then moving to self-play on GPU clusters.
On the research side, I worked on causal feature selection to figure out which features actually cause the outcome by combining TabPFN embeddings with neural Markov blanket predictors. I also built a multi-agent code generation system where two RL agents collaborate to write better code than one. Outside of ML, I built Athena to learn iOS development. It generates quiz questions from your docs and sends them as notifications throughout the day.
Won first place in the Sustainability track at HackHarvard '24 using satellite imagery to detect urban heat islands and surface actionable planning insights. I was also one of 18 selected from ASU out of 2,512 applicants for a DAAD RISE research grant from the German Academic Exchange Service.