Roles & projects

Joined as Lead Backend Engineer and became Head of Engineering, growing the team from 3 people to 15. Worked with product and go-to-market teams to plan, architect, and deliver all engineering work, and wrote PRDs for new features.

Built hiring, onboarding, and management practices from scratch. Owned relationships with key hardware and software partners and worked with Customer Success to triage, debug, prioritize, and fix customer issues. Joined enterprise sales cycles as both engineer and head of security, and helped close large deals.

Set engineering standards by example as well as by management. I built major backend and Android systems and full-stack features myself.

• TypeScript
• Node.js
• Android
• Kotlin
• GCP

Led technical direction for a critical component of the Search Ranking backend. Reduced recall loss by more than 10% through large-scale ETL pipelines, feature engineering, and compression and performance work, and helped other teams run similar projects.

Redesigned the deep neural network in the critical serving path, cutting tail search latency by 15ms. Became the team's go-to person for deep learning and ran an internal AI reading group. Built and maintained the testing infrastructure for a major infrastructure redesign, including coordinating on-call around it.

Other work: ranking and retrieval for Google Maps Events that doubled CTR and DAU; machine learning ranking signals for local search quality on Search and Maps; and a 20% project with the TensorFlow Object Detection team using CNNs to identify tree species in major cities from satellite and street-level imagery.

• Machine Learning
• TensorFlow
• ETL
• C++

Human-Computer Interaction track. 1st place, CS Department Software Fair 2017, for disaster response software. Stanford Debate Society member 2013–2017, captain 2014–2016.

Led ManageXR’s move to agentic software development, which meant changing how the team plans, ships, and reviews work. Introduced engineers to Cursor and Claude Code, gave internal talks on AI-first development, and helped the business side use AI and MCP tooling for customer support and churn tracking.

Most of the work was in the surrounding system. Getting agents to do real end-to-end work required fast, reliable tests they could run to check their own changes; clear limits on what they could touch (local edits, dev, anything close to prod); and AGENTS.md-style documentation the team updated after a session went wrong, so the same mistake didn’t happen twice.

The result: engineers regularly shipped in hours what used to take days. For the fastest engineers, QA became the bottleneck instead of the agents.

• Claude Code
• Agentic workflows
• Process design
• Tooling

Designed and led security and compliance at ManageXR, from initial policies and controls through annual re-certification. Took the company through SOC 2 Type II and ISO 27001 certification, which enterprise deals required, and built the audit and IT operations process that keeps both current as routine work.

• Security
• Compliance
• IT Operations