Education
Stanford University, Sep '16 - Jun '18Master of Science in Computer Science
GPA 3.90
Rice University, Aug '12 - May '16
Bachelor of Science in Electrical Engineering
GPA 3.90, Cum Laude, Distinction in Research and Creative Work
Professional Experience
XMotors.ai, Santa Clara, CASenior Behavior Planning Engineer, Jul '23 - Present
- Design and implement state-of-art behavior planning and motion planning algorithms for autonomous driving system, including urban-driving, highway-driving, and driving in unstructured areas
- Design and implement behavior planning algorithms and data structures foundational to the autonomy stack, specifically developing the traffic light module under standard-definition map
- Work closely with other teams to research and identify improvements for compute platforms, middleware libraries, and infrastructure pipelines, including provide prompt bug fixes and interface upgrades
- Conduct major system development, design, and testing as a leading engineer in the team, in order to analyze and/or improve current business logic and algorithms
Robotics Engineer, May '18 - Jul '23
- Design, develop frameworks and optimize algorithms for robot control and motion planning in C++
- Implement trajectory tracking methods in C++ for fast, accurate, and smooth robot motion, full stack from velocity profiling to driver control
- Implement collision checking/avoidance methods in C++ to ensure safe robot movements
- Develop toolchains for robot simulation with physics engine, as a substitution of ROS
- Integrate perception models, control stack, and peripheral sensor modules to build robot applications
- Design workflow and business logic of a line-kitting application, implemented in Python and deployed onsite for a major client
Research Assistant, Stanford Vision Lab, Dec '16 - Jun '18
- Set up robotics perception and learning infrastructure, including future Stanford Arm Farm
- Deployed virtual reality interface for scaling up task demonstration data collection
- Conducting research on imitation learning and RL under context of robotics
- Conducting research on robotics sim-to-real transfer learning problem
- Hold 2-hour session every week, grade homework, make exams and lecture review sessions.
Software Developer Intern, May '17 - Sept '17
- Implemented cloud service usage prediction model with deeplearning4j
Olympic Weightlifting Specialist, Personal Trainer Group, Jun '16 - Aug '16
- Olymipc lifting training sessions--snatch, clean and jerk
Research Assistant, Rice Efficient Computing Group, May '15 - July '16
- Built image-classification energy reduction simulation framework RedEye based on Caffe
- Extended GoogLeNet Deep Convolutional Network with noise layer and energy cost layer
- Trained ImageNet data on modified Caffe framework for optimal noise-energy trade-off
- Designed RedEye chip memory layout on Cadence
- Developed stride/motion detector that recognizes, analyzes motion and generates feedback
- Integrated multiple accelerometers, pressure sensors, and SD cards on MSP430F6726
- Conducted Bluetooth-based real-time data processing in MATLAB
- Conducted DSP on gathered sensor data, including feature extraction/model matching
- Designed, tested, and revised motion recognition algorithms based on gathered sensor data
Fundamentals of Electrical Engineering, Multivariable Calculus, Single Variable Calculus
- Graded homework and evaluated students’ performance
- Held TA sessions 3-4 hours every week to answer students’ questions
Consultant, Mar '12 - Aug '12
- Taught two classes of high school students 3-hour lessons on SAT reading weekly
- Provided personal consultation for five students on college pitching, essay writing, and application logistics
- Benchmarked services with other educational consulting companies and reported possible improvements
Selected Projects
Cloud-Crowdsourcing Robotics, Mar '17 - Jun '18- Developed keyboard, VR, phone interface for teleoperating robot in simulation
- Devised and designed setup for daily tasks that are scalable for crowdsourcing
- Implemented task demonstration environment in simulation with Bullet Physics
- Performed data mining on 1.1 billion taxi records by AWS EMR cloud computing
- Used PySpark for massive data processing and information extraction
- Implemented MDP value iteration to generate optimal passenger pick-up policies
- Visualized data and simulation paths with tools such as gmplot
- Attached dropout layers on RGB-depth Convolutional Neural Network during test phase
- Generate qualitative and quantitative analysis on predicted depth maps
- Designed and built a high-throughput, low-power wireless alternative to current Electrocochleography (ECoG) technologies. The system includes 16 neural electrodes that collects the signal, an analog front-end Intan RHD2132 that samples the signal, a micro-processor Nordic NRF52 that compresses and sends data, and a secure server that displays received data in real-time
- Designed Use Case Diagram and system block diagram for API construction
- Used Java Remote Method Invocation (RMI) for data package exchange
- Applied multiple layers of MVC to encapsulate chat rooms that support text and file transmission
- Parsed a set of documents into a vector-based corpus database for Gibbs sampling
- Implemented an M-Tree to store parsed documents
- Implemented the query system that compares input document with the corpus and displays top K most similar documents
- Assembled motor driver SN754410 and luminosity/ultrasonic sensors for obstacle detection
- Designed and installed PCB with amplifiers, voltage regulators, and sensors
- Implemented PWM for motor control and used I2C protocol for ADC module
- Integrated NeoPixel LEDs and micro-speaker for barrier-range display and alarm
Publications
RoboTurk: A Crowdsourced Platformfor Robotic Skill Learning through Imitation, CORL 2018
Ajay Mandlekar, Yuke Zhu, Animesh Garg, Jonathan Booher, Max Spero, Albert Tung, Julian Gao, John Emmons, Anchit Gupta, Emre Orbay, Li Fei-Fei, Silvio Savarese
Neural Task Programming:
Learning to Generalize Across Hierarchical Tasks, ICRA 2018
Danfei Xu, Suraj Nair, Yuke Zhu, Julian Gao, Animesh Garg, Li Fei-Fei, Silvio Savarese
RedEye: Analog ConvNet Image Sensor Architecture
for Continuous Mobile Vision, ISCA 2016
Robert LiKamWa, Yunhui Hou, Julian Gao, Mia Polansky, Lin Zhong
Patents
Robotic Multi-Item Type Palletizing & Depalletizing, US_10,549,928_V1, Feb. 4, 20'Robotic Handling of Soft Products in Non-Rigid Packaging, No. 16/797,359, In-Process, Feb. 21, 20'
Robotic Carriage Tray Table, No. 17/960,544, In-Process, Oct. 5, 22'
Emulated Robot Peripherals, No. 63/435,176, In-Process, Dec. 23, 22'
Multiple Robot Simultaneous and Synchronous Pick and Place, No. 18/393,504, In-Process, Dec. 21, 23'
Selected Honors
- IEEE Eta Kappa Nu Rice Chapter (IEEE-HKN)
The honor society of IEEE dedicated to recognizing individual excellence - Louis J. Walsh Scholarship in Engineering
$1,500 scholarship for deserving engineering students at Rice University - President’s Honor Roll
Recognition of undergraduate outstanding academic achievement - Distinction in Research and Creative Work
Awarded to at most 10% of graduating seniors - Bill Wilson Best Senior Design Award
$250 prize for Rice University Affiliates Day senior design showcase