Jia-Chi Wu Email & Phone Number

San Jose, Ca, Us

Santa Clara, Ca, Us

Austin, Texas, Us

- Autopilot replay (aka open-loop simulation) system tech lead- Enabled running current-gen heterogeneous-computing Autopilot stack on the expanding next-gen hardware farm for replay, greatly reducing turnaround time of replay and intermediate training data generation- Eliminated Autopilot full-stack behavior nondeterminism in replay by inspecting and improving heterogeneous-computing and multi-SoC/multi-process/multi-threaded code in Autopilot, making replay an indispensable Autopilot behavior debug tool

Cupertino, California, Us

- Designed and developed C++ deep-learning inference library capable of switching and mixing CPU/GPU/Bionic processing; library was used extensively during Face ID development- Productized and shipped embedded CPU core portions of deep learning-based face detector and attention detector for iPhone X Face ID- Productized and shipped embedded CPU core portions of deep learning-based face detector and tracker for iPhone XS as replacement for 3rd-party licensed HW detector- Designed, implemented, and delivered camera abstraction framework with both C++ and Swift API to provide company-wide early access to the iOS LiDAR camera- Proof-of-concept prototype to use automatic code generation to wrap and export C++ code as TensorFlow custom kernels for use in Python

New York, Us

- Worked regularly on code involving 3D math, computer vision, linear algebra, and computer graphics- Optimized multi-threaded code and compute shaders and improved tracking frame rate from 60 FPS to 120- Resolved long-standing and previously undetected bugs in lens calibration, which enabled the use of more affordable lenses, resulting in reduced costs- Developed a method to detect walk cycles and track users without requiring a calibration pose, maximizing the number of gait cycles in tracked motion for gait analysis- Developed an optimization-based algorithm that resizes skeletons to fit users of different body shapes, improving the quality of tracking

- Surveyed, evaluated, and implemented ideas from research papers for problems such as vision-based hand shape classification and alpha matting- Collaborated with the engineering team to transform a motion-gaming research project into a commercial software tool for game development- Implemented an active-ragdoll system that uses players’ motion from Kinect, pre-recorded motion, and physics simulation to synthesize new animation

Redmond, Washington, Us

- Developed a photo sharing and annotating applet for the (then) next-generation OS using P2P technology. Development was done using a presentation engine that was still under heavy development and rapid change.- Represented the group to compete in a company-wide applet contest with many other full-time developers, demonstrated functionality of the applet, and won the 1st prize.

Redmond, Washington, Us

- Constructed reusable UI components for document preview with adaptive and dynamically changing layouts (United States Patent 7444598).- Built prototype applications for browsing and previewing of Word/PowerPoint file contents and e-mails.

Doctor Of Philosophy (Ph.D.), Computer Science

Bachelor'S Degree, Computer Science