- Head TA for CS61A, a 2000 student second-look CS class- Lead team of 9- Large scale architecture design, run scrum/agile- Building and maintaining tools used by entire EECS department- ML (LLMs + RL) + convex optimization- Primary technologies used: GCP, Flask, React, Jira, Cloud SQL, DockerMore details upon request.

- TA for CS61A- Teach discussion and lab (~40 students)- Hold technical office hours- Write software to maintain and update infrastructure used by all UC Berkeley EECS lower division classes (technologies used: React, Flask, SQL)

- Course tutor for CS61A – the Structure and Interpretation of Computer Programs, a class that services 2000 students- Expanded software infrastructure with React, Flask, Node.js- Run small-group tutoring sessions teaching and reviewing the course’s content- Held office hours and homework parties, answered student questions- Help proctor and grade exams at scale- Work alongside 90 other course staff members

- AI4Science in the Ceder Group (electrochemistry)- Working on deep learning for accelerated materials discovery- Developed algorithm for uncertainty estimation on graph neural networks- Building ultra-large scale, distributed inference platform for graph-based interatomic potentials- Linux, PyTorch, DGL- Python, Cython, C, OpenMPWe published in Nature Machine Intelligence! https://www.nature.com/articles/s42256-023-00716-3

- We discovered a novel electrochemical technique to convert carbon dioxide and waste product into valuable industrial feedstock- The end goal is to provide an economically viable carbon sink for large scale carbon emitters - Fundraising, competitive analysis, tech leadership- Left as my interests diverged towards machine learning for science applications- My cofounders will do great things with the company

- Worked on NLP research applications for the Commercial Bank AI team- Performed literature reviews on state-of-the-art NER, contrastive learning, and federated learning techniques- Designed contrastively learned architecture for inferring company relationships from a corpus of news text- Proposed and implemented a highly parallelized framework for boosting NER model performance on a new distribution of data in an unsupervised fashion (framework boosted consistently performance by 20-30%)- The framework was a novel curriculum learning regime involving several models "supervising" a single model

Technologies/Languages used: Swift, Python, .Net Core, Razor Pages, Objective-C, ASP.NET Identity, Azure App Services, EF Core, SQL, Azure Functions, Cocoapods, Stripe, RESTful APIs, Chart.js, Video.js

- Working on team creating diagnostic device centered on functionalized carbon nanotube forests in order to rapidly and accurately detect airborne and liquid-based biomarkers (including COVID-19 molecules)- Expanded software backend for the device, focusing on backend-board communication- Created custom testing framework and wrote integration and unit tests with JUnit- Written in Java

I created a cross platform app (IOS and Android) that locally stores and displays the Region IX Standard Operating Procedures - Northwest Community EMS Edition as well as the COVID19 supplemental Standard Operating Procedures. The app is used by the Northwest Community Emergency Medical Services System's (NWCEMSS) first responders and firefighters. This app has also incorporated in their EMT training programs, and is in use at 21 fire departments, 3 ambulance companies, and 5 hospitals.