I love making (in its broadest definition). My background is at the intersection of Electrical Engineering and Computer Science (focused in AI).

I currently lead a couple projects as a senior ML scientist at Cerebras. My work focuses on the training and deployment of LLMs:

• Applied Research: Domain adaptation, Information Retrieval, Post-training, Inference time compute
• Engineering: Agentic & RAG applications, wrote much of the basis of Cerebras's LLM model library, and many many fun engineering tools (both internal and external)

Prior to that, I co-founded and served as the CTO of Agile Data Driven Decisions (AD3) for four years. AD3 works with enterprise customers to optimize their global supply chain in order to not only make them more efficient but also more resilient. During my time, I led our team through various technical challenges ranging from creating machine learning models for high stakes decisions ($10s M on the line) to building low-level highly parallelizable supply chain simulators. As with any small startup, I got to work on much more than just engineering: strategy, pitch decks, sales calls, hiring, etc. You can read more about our work in this Harvard Buessiness Review article: How Machine Learning Will Transform Supply Chain Management

During my time at Stanford, I got to explore three academic interests in depth:

• I wrote a series of papers (culminating in my honors thesis) at the intersection of signal processing, music, and machine learning. A large portion of the work focused on using deep learning for audio source separation (I was incredibly interested in creating tools for musicians to sample music). I was honored to receive the Stanford EE design project award for one of my projects.
• I was interested in Large Language Models (which at the time were actually quite small). One of my works, "Recursive Transformer" stumbled upon the same intuitions and ideas as chain-of-thought and inference-time-compute (now popular techniques in the field).
• I also spent a lot of time working on low-level software and hardware architecture. Some of my favorite class projects included designing a custom ASIC for neural networks, building a distributed consensus system (RAFT) from scratch, and writing an async I/O library in C from scratch (which I later found out was implemented similarly in Facebook's internal systems).

Much of where I am today is attributed to my early exposure to engineering. In high school, I converted part of my room into a makeshift electronics lab. Some of my favorite things I built included a drone, a pulse monitor, a traffic counter for smart homes, and a low-cost collision avoidance system for airplanes. I was lucky to have much of this work recognized by the science fairs run by Google, Intel, Regeneron, etc. and was honored to be part of the Center for Excellence in Education's Research Science Institute program.

Aside from tinkering with circuits and coding, I also enjoy creating digital art, playing the bass, or building Lego art.