• implemented AI Search feature across project documents and multiple DB models

• integrated AWS Kendra to index and search documents for user query

• integrated embeddings and vector search into DB tables to allow search across DB tables

• reduced AI Search latency from initial 30s to ~5s using Logfire to pinpoint bottlenecks

Worked on a suite of fullstack projects including:

• database migration of billions of flashcard records from MySQL to microservice backed by Google Spanner (PHP, Kotlin, Java)

• added improvements to Checkpoint game microservice (node.js, websockets, redis, react/redux)

• increased usage of premium study features with Test-To-Learn as Tech Lead (react/redux, Kotlin, Hack)

• added improvements and bug fixes to unstable Achievements Service (Kotlin, Firestore)

Have experience with Sprint and Kanban styles as well as experience with JIRA, Rollbar, CircleCI, ArgoCD, Github.

Backend Engineer on Driver Excellence

• Tech Lead for Feedback Home - a feature that summarizes driver performance in-app. Lead to a ~7% reduction in driver suspensions and lower support costs

• Pushed against management to prioritize stability work over roadmap work -> over the course of 2 months, I reduced the number of pages plaguing our oncall by 60% and fixed a feature that saved Lyft 40k rides weekly

• Lead of New Year's Eve readiness work -> coordinated a team of 4 engineers to divide and conquer stability work needed for the busiest night of the year for Lyft -> lead to an error-free NYE for our team

Backend Engineer on Pricing

• Sole backend engineer on data visualization tool for Lyft data scientists leading multiple through product cycle (requirements gathering -> spec'ing -> feature implementation)

• Built custom data pipeline to move data from S3 into AWS RDS - optimized pipeline by streaming files line-by-line and batching writes bringing migration time down from ~30 minutes to ~5 minutes.

• Took ownership over previously owner-less Helium service (the backend for many data vis tools). Since then, test coverage in the service has increased 15% because of my tests alone, and I was the first one to put stats and alarms on the service.

Teach a pod of 10-20 soon-to-be engineers the fundamentals of programming and monitor their performance to recommend them for the full curriculum.

• Self-taught PRAAT scripting language over holiday break to more quickly analyze speaker data. I was brought onto two additional projects because of it

• Worked overtime after computer malfunction to recover lost data -- recovered ~80% of data

• Was called one of the most versatile and useful assistants by researcher