Featured Work
Top Project
Pixel Watch Launch
Pixel Watch 4 Gestures
I architected and launched the One-Handed Gestures (Double Pinch & Wrist Turn) for the Pixel Watch. This involved implementing low-power ML models using C++ and TFLite, achieving >92% accuracy with sub-1ms inference time.
Experience
Google
Sep 2022 – Present
Software Engineer
• Mountain View, CA
WatchSW AI/ML & Algorithms
- Launched One-Handed Gestures (Double Pinch & Wrist Turn), a novel, low-power ML gesture recognition algorithm (C++, Python, TFLite) for the Pixel Watch, partnering with a Research Scientist to lead end-to-end integration, on-device optimization, and sensor fusion (IMU & physiological signals); achieved over 92% accuracy, 1ms inference, and <250ms E2E latency.
- Architected the Gesture Hardware Abstraction Layer (HAL) to ingest and transform sensor events for the platform service, decoupling the algorithm to enable vendor-agnostic implementations (C++); introduced bidirectional communication for context-aware sensing (e.g., display state), facilitating cross-device orchestration with xDevices such as XR glasses and reducing latency by 30% through the consolidation of abstraction layers.
- Owned and optimized the Low Latency Off-body Detection (LLOB) algorithm – a core component for most algorithms and system performance – resolving false detections and power regressions, raising reliability across all shipped devices (C++, Matlab)
- Implemented on-device telemetry and ETL dashboards (Python, SQL); led Gestures and LLOB bug triage, closing 1000+ bugs.
- Mentored teammates on sensor integration and MCU-level development; led design reviews and authored the "MCU-based Practical Telemetry Guide" referenced by 24+ engineers.
Cloud Asset Inventory & Search
- Productionized a scalable C++ solution to enrich Cloud Asset Inventory with structured metadata for cloud-managed assets, impacting 90%+ of GCP resources and boosting adoption to 500,000+ active users.
- Enabled launch of Asset Enrichment in the Asset Query System (C++, SQL, Spanner), powering efficient queries across 275+ GCP asset types for SCC customers.
- Drove cross-team optimization of a workflow runner (gRPC, Borg, Python), reducing development time by 25% resulting in 10 SWE weeks saved.
OpenSesame
Jun 2020 – Aug 2020
Software Engineering Intern
• Portland, OR
Front-End Division
- Enhanced course language selection with searchable dropdowns (Angular, TypeScript); built and ported 10+ end-to-end tests from Drupal to Angular/Selenium, expanding QA coverage.
- Resolved 6+ sprint bugs—including a critical IE landing page issue—by extrapolating burn-down processes and strengthening test automation with Behat API, and unit tests.
ANDSystems
Jun 2019 – Aug 2019
Machine Learning Intern
• Ulaanbaatar, Mongolia
Machine Learning Team
- Analyzed the purchase history of 100,000+ users buying coupons by regression analysis using Python, Pandas, NumPy, and PyTorch, identifying an under-marketed sector in sales that increased revenue by 10%.
- Launched an MVP module-based recommender system with caching for an e-commerce platform (BananaMall) with 10,000+ downloads on the Playstore and 100,000+ users, using Python, SKLearn, and DynamoDB.
University of Rochester
Sep 2019 – Dec 2021
Teaching Assistant & CETL Tutor
• Rochester, NY
Computer Science Department
- Taught students in Web Programming, Data Structures & Algorithms, and Formal System & Computations courses.
- Conducted tutoring sessions with 15+ college students, resulting in a 30% increase in their respective course grades.