Patrick Kao Email & Phone Number

Boston, MA, US

Cambridge, MA

-Implementing DESIRE trajectory prediction deep neural network in PyTorch to provide lab baseline accuracy metric-Preprocessing data using NumPy, programming RNN encoder-decoder, CVAE, CNN, and feature-pooling modules, training neural network, and tuning hyperparameters to improve trajectory prediction performance-Developed Jenkins integration test.

Menlo Park, CA

- Optimized CNN video feature aggregation strategy to enable large-scale copy detection and free over 1000 machines used for legacy system- Curated datasets with SQL and used Pandas to preprocess datasets for Pytorch models

Menlo Park, CA

- Wrote XCTest UI tests to verify Intune MAM policy behaviors- Integrated UI tests into Azure Pipelines CI/CD system, automating regression detection

Santa Clara, California, United States

-Converted existing CMake build framework to Bazel for autonomous vehicle platform code, improving modularity and decreasing build times for downstream customers-Wrote Bazel packaging rules and integrated them into CI/CD pipeline, automating packaging and enabling code delivery

Cambridge, Massachusetts, United States

-Automated building and Docker containerization of Windows to Linux application translation tool-Integrated automatic tool into downstream CI/CD pipelines, eliminating manual upkeep requirement, improving developer efficiency, and increasing build reliability for customer-facing applications

San Jose, CA

-Developed machine-learning-based trajectory maneuver classification system and applied it to organize company’s entire data corpus-Automated existing data generation pipeline with Airflow and pioneered Jenkins regression test system

Cambridge, MA

-Worked on sponsored project to design and prototype drones that function in extreme environments-Implemented control, navigation, visual odometry, and sensor processing software

San Jose, CA

-Engineered mobile robots capable of automating company map data collection-Designed hardware configuration and programmed software stack capable of autonomous control, localization, and object recognition

Stanford, CA

-Defined specifications and built hardware for a quadcopter capable of carrying heavy loads-Evaluated the suitability of visual odometry and Simultaneous Localization and Mapping (SLAM) software packages in C++ and Python for tracking the position of a quadcopter using an onboard camera-Presented work to Professor Marco Pavone's ASL Lab and AA 290 masters.

Stanford, CA

-Mentored SUAVE 101 course and taught Stanford students how to build hardware and configure software to fly RC quadcopters and airplanes-Prepared new prototype aircraft for SUAVE 101 course and wrote instruction manual for students-Invited as a SUAVE presenter at the 2016 Stanford Aeronautics & Astronautics School of Engineering Affiliates Day Poster Session

Stanford, CA

-Collaborated with team to try to build the world's most powerful quadcopter and break the world record for fastest climb to 100 meters (Guinness World Records application in progress)-Sole programmer and wrote the flight control software in C++ to stabilize high power quadcopter-Independent work on flight control software won Finalist and 1st place.

Master Of Science - Ms, Computer Science

Bachelor Of Science - Bs, Computer Science, Gpa: 5.0/5.0

Activities and Societies: Design Build Fly, MIT Asian Dance Team, Projx, Zeta Beta Tau

Gpa: 4.366/4.0

Awards: National Merit Finalist, Rensselaer Medal (given to top STEM student in class of 400+)