Tech + business work with startups

Analytical layer for alumni data

- Developed Federated Learning test environments and created a novel backdoor threat model called Double Momentum Backdoor Attack and combining Di erential Privacy and Anomaly Detection to create defense mechanisms. Average ~40% in backdoor task accuracy than baseline models- Published paper on “Double Momentum Backdoor Attack in Federated Learning” in the Journal for Student Research- Published papers on “Noise-Resilient Federated Learning Models through Robust Aggregation Methods” and “LewyScan – A deep learningapproach to diagnosis between Dementia with Lewy Bodies and Alzheimer’s” in the SigmaXi Student Research Conference

- Built the minimum viable product for a two-sided marketplace for neighbors to rent and lend essentials. Launch the web platform using Next.JS, GraphQL, PostgreSQL, Tensor ow, Google Cloud Platform, and Swift (for the iOS prototype).- Launched a weekend sale that received $300+ in revenue from customers in 4 di erent markets. Awarded the “Most Innovative Project” in the MIT LaunchX program

Worked with Ph.D. Candidate Negin Majidi on a novel countermeasure against adversarial agents in Federated Machine Learning systems in high-noise and backdoor dataset environments. Developed research direction; designed, programmed, and executed experiments; and presented findings to the broader UCSC community

Organized a global virtual weeklong startup hackathon in July 2020 and hosted a couple of fireside chats with many entrepreneurs and CEO, including Astro Teller (CEO of GoogleX), David H. Hansson (Creator of Ruby on Rails), Divya Jain (Director of Adobe Sensei), and many more.