Took leave from Lyft to attend a self-directed educational retreat for programmers. Focus areas: Rust, computer systems, programming languages. Selected projects: ‣ Mini-LSM: Toy implementation of an LSM-backed key-value store. ‣ Rust Toy C Compiler: Toy compiler for a very small subset of C to AArch64 assembly. ‣ LLM Says Buttons: Button-based interface for LLM interaction encouraging more structured conversations.

Oct 2022 - Mar 2025: Fulfillment Offer Generation Tech Lead for Cost Pod (since May 2024). Led staff + senior engineers across multiple teams in a year+ long effort to integrate ETA, Driver Pay signals into cost prediction Airflow + Spark workflows, unblocking product launches across Marketplace Org. Optimized Spark pipelines decreasing runtime of cost prediction workflows from 10+ hours to <1 hour. Implemented new signal for driver location opportunity cost projected to increase annual PMM by $12M. Required multiple layers of caching and NumPy vectorization to overcome latency cost of network hops. Aug 2019 - Oct 2022: Pay Platform Led year-long initiative to build Lyft-wide subscription billing platform decoupling recurring billing management from Memberships services. Unlocked new pay features resulting in 24% decrease in recurring payment failures. Implemented new refund flow during migration of transactions database and data model from NoSQL (MongoDB) to Aurora RDBMS. Built refund tooling to resolve customer tickets and perform incident remediation.

Built a new Apache Flink stream allowing input of road closure information into Lyft’s real-time speed observation system, increasing accuracy of ETA and route calculation. Wrote a new microservice pulling third-party traffic incident information into the Lyft traffic pipeline.

Designed and implemented grammar-checking logic for alpha version of Perfect Tense, a grammar correction software project. Contributed 50% of grammar-checking code and 15% of overall project code at time of alpha release.