ML Infrastructure - Training

• Exponentially grew customer base by 800% by leading the design, development, and deployment of critical features.• Decreased error rates in LLM generations from 40% to 3% across short-form and long-form copy through different methods of prompt engineering, output constraining/chaining, and hallucination detection and remediation. • Improved system uptime by 10x and latency by up to 20x through the implementation of a distributed backend and parallelization.• Increased engineering velocity by 2x by leading contract engineers and designers.• Shipped rapid-prototype features on timelines of < 1 day and full features on timelines of < 3 days.• Created CI/CD practices for engineering team to improve system reliability.• Introduced Linear as a project management tool to increase transparency and promote open-air documentation.• Fostered a culture of swift and intentional iteration through holistic documentation, user-testing, and dog-fooding.

• Researched the synthesis of sequential programs into distributed programs by leveraging CRDTs.• Worked on HYDRO project, developing cloud-native programming models for scalable and resilient distributed applications.

• Improved reliability and safety of 10,000+ machines performing 100,000+ jobs by engineering an Atomic solution for a Google Borg safety mechanism.• Surfaced and aggregated 15+ key Google Borg metrics into a customizable dashboard for the Borg SRE team.• Designed and implemented 3 new metrics to enhance service diagnostics.

• Source and create opportunities for active members to develop their professional skills and be more involved in the Berkeley startup ecosystem.• Created a speaker series featuring founders.

• Taught practical business skills used by startups, culminating in a capstone venture project which is pitched to a group of 30+ investors and those familiar with the startup ecosystem.

• Developed dev-ops tool used by 1000+ Data Engineers for rapid deployment of data analysis and machine learning models.• Wrote a common line interface to generate custom Python libraries for AWS data consumption and production.• Automated secrets handling and distributed systems setup for data processing and ML jobs.

• Developed over 200+ UI components for company-wide design system.• Increased engineering velocity by 1.5x by establishing CI/CD platform for frontend development• Created web application for Parrot, a $3m ARR solution for collecting SMS payments on behalf of SMBs.

• Connected 2559 participants from 100+ countries, with 80+ mentors, 100+ judges, and 23 sponsors to produce 224 project at hack:now.• Managed $275,000 in funds sourced from 40+ sponsors after Cal Hacks 6.0, seeing over 2000 participants and 3000 applicants.• Launching hack:month, providing a community for 100+ students to develop their side projects.• Formulated 3 weeks worth of web development curriculum for CubStart, educating 90+ Berkeley students.

• Organized the judging process at Cal Hacks 6.0, which coordinated over 60 judges to interact with over 200 projects in under 3 hours, all while providing invaluable feedback.

• Taught a group of 30 students topics including linked lists, asymptotic analysis, higher order functions, trees and traversals, persistence, hashing, heaps, priority queues, graphs, graph algorithms, and sorting algorithms.• Mentored for 15 labs, 4 projects, and 3 exams.

• Built a front-end website algorithm using JavaScript currently applied on over 200 online news articles.• Optimized data management from backend to reduce extraneous operations by 40%.