Drove the execution of a multi-phase migration of private pricing contract data, reconciling different data models while preserving downstream billing compatibility. Designed and implemented system retrofits to maintain legacy contract semantics, executed large-scale backfills via MCM protocols with layered audits, and authored system-wide validations to ensure data integrity ahead of billing cycles. Partnered closely with downstream consumers during billing to mitigate misbilling risk.

Built reporting and operational tooling (ops reports, billing dashboards, validation scripts) to support audit accuracy and operational visibility.

Earlier, owned a region deployment of a billing-critical service to an air-gapped partition, involving implementing environment-specific configurations, security-compliant workarounds, validation, and operational readiness to support isolated production environments.

In parallel, picked up operational support for the billing system, including oncall rotations and monthly billing cycle support. Strengthened team operational processes by authoring SOPs, updating runbooks, and building automated validations and audits to improve billing reliability.

Served as a teaching assistant for Caltech's advanced robotics course, supporting student teams in designing, building, and programming custom robotic arms (5–7 degrees of freedom) capable of fully autonomous operation. Guided teams through the integration of hardware, kinematics and control, computer vision, and autonomous behaviors using ROS (Robot Operating System). Assisted with debugging hardware and software issues, provided feedback on team designs, and helped teams scope, prioritize, and triage features in preparation of final demonstration.

Served as a teaching assistant for Caltech's core computing systems course, supporting students in bridging the gap between high-level software and underlying hardware. Guided students through topics spanning computer architecture (x86-64 assembly), memory management (virtual memory, caching, dynamic allocation via custom malloc, garbage collection), process management and concurrency (processes, threads, pthreads, mutexes, condition variables, signals, exceptions), and systems security (buffer overflow exploits, side-channel attacks, Meltdown). Assisted with grading and extensive debugging office hours across projects including a simplified Java Virtual Machine, a binary bomb, a BASIC-to-x86-64 compiler, custom memory allocators, a virtual memory OS kernel, a real-world Meltdown attack implementation, and a concurrent password cracker.

Developed a graph analysis tool for Amazon's retail ordering orchestration platform, which managed large-scale workflows (hundreds of nodes, thousands of edges) coordinating business logic from purchase through delivery. Leveraged production logs to trace node entry/exit patterns and identify workflow optimization opportunities (for example, unreachable nodes, never-traversed edges, and consistently failing states). Analysis identified 15–25% reducible nodes and up to 30% reducible edges across workflows.

Designed and built the tool's internal UI console, enabling engineers to explore workflow analysis results interactively.

• Wrote front-end for contactless payment startup

• Gained experience with rapid-iteration startup work

• Learned about routing, Firestore databases, hooks, and asynchronous programming