Organization: Search & Conversational Shopping

• Managing proactive/reactive guardrails for Amazon Stores LLMs using low-latency string processing and NLP techniques

• Finished a 2-quarter system redesign to simplify data model, enforce data consistency, and save 2 mo/yr in effort, with no outages

Organization: Fulfillment Optimization

• Built mission-critical tooling and workflows to explain fulfillment control systems that use linear programming and ML

• Led 3 devs end-to-end to build a simulation system to enable rapid iteration on controllers' heuristics and hyper-parameters

• Mentored 2 interns through doc writing and backend/UI development on operational and engineering excellence projects

• Onboarded to and unblocked 2 sister teams' critical-path projects ahead of schedule by doing "away team" work

Organization: Fulfillment Optimization

• Developed shipping method evaluation and ranking features for one of Amazon's most critical and highest-scale systems

• Delivered a new system to replan risky shipments using big data processing, an event-driven service, and React, saving $2MM/yr

• Saved $5.3MM/yr in infrastructure costs via host type migration, JVM/GC tuning, and autoscaling with no performance drop

• Successfully improved system resiliency score by more than 3x in time for Prime Day as part of a VP-level campaign

• Assisted faculty in teaching CS24 (intermediate C++) and CS185 (human-computer interaction)

• Answered 100s of student questions, taught weekly sections and labs, held weekly office hours, read and evaluated student work (homework, programming assignments, exams), and proctored exams

• In CS185, I mentored 5 teams of 5 students each on their term-long design projects, which involved needfinding, ideation, low- and high-fidelity prototyping (Figma), and prototype evaluation

• Reduced manual effort to analyze transportation capacity breaches during shipping by 75% compared to prior tools and SOPs

• Implemented a Ruby on Rails dashboard for 100+ transportation operations users to visualize capacity breach root causes

• Developed a Java service to build capacity breach analysis queries from user input and run jobs on Amazon Redshift

• Designed and implemented a novel concolic testing search heuristic based on branch selectivity

• Extended CREST, a concolic test generation tool for C programs, to support the heuristic