I’m Satwik, a Computer Science & Data Science student at UC Berkeley. I am currently working on building World Models to build out ontologies in non-visual domains
In the past, I’ve worked on a variety of projects (from most recent to earliest):
- I was an early engineer & scientist at GRU Space where I worked on building Reinforcement Learning models for lunar regolith mining, as well as, built out specs for mineral prospecting satellites and lunar gondolas.
- I was a founding engineer at LawLoop where I built proactive AI agents that helped law firms complete case research and other back office tasks without any prompting needed
- I was the youngest engineer at Advocate where I built most of the ML infra for their LLM-facing insurance compliance software that is integrated into their core platform.
- Researched under the Professor Prasant Mohapatra on the topic of Double Momentum Backdoor Attacks in Federated Learning.
- Worked on Noise-Resilient Federated Learning Models through Robust Aggregation Methods at the UC Santa Cruz SIP Program. Presented at the Sigma Xi Research Conference.
- Developed a multimodal extractor that uses NLP and Computer Vision to extract information from Mental Health Records, MRI, and other biomarkers to differentiate between Lewy Body Dementia and Alzheimer’s Disease.
- Built a startup called Neigh — a two-sided marketplace for neighbors to rent and lend essentials to one another.
I’m also interested in DeFi and Ethereum Development (check out my analysis on the 3Jane Protocol) at Blockchain@Berkeley.
I think values are important thus, here is my Code of Life.
On the side:
- I’m a massive Warriors fan (Go Dubs!)
- I love to read and write about practically anything
- I’m big into working out and staying fit
- I’m a conisseur of wildly vast musical genres
I’ve made it my goal to build and learn in public which is what I intend to do here. If you have any interesting recommendations, want to work on something together, or just want to meet reach out through email or twitter @therealsatwik!