Kun Yang Email & Phone Number

Cupertino, Ca, Us

Sunnyvale, Ca, Us

Charlottesville, Va, Us

Santa Clara, California, Us

1. Implemented an interface enabling message transfer and remote control between python-based deep learning framework and NS3 network simulator.2. Enabled a reinforcement learning algorithm that can effectively handle a radio resource management problem in a network-slicing scenario for delay-SLA violation minimization. It improved the violation rate from 40% to 10%.3. Developed an API that enables the remote control between the openfl and the NS3, enabling federated learning with more realistic network simulation.

Guangzhou, Guangdong, Cn

◦ Re-implemented well-known vehicle behavior prediction models VectorNet and TNT on an in-house dataset and delivered simulator-based demo.◦ Investigated effectiveness of different feature encoding methods in a vehicle behavior prediction system◦ Designed a new scheme combining VectorNet and Residual network, reaching 80% of TNT’s performance using 60% of the parameters

San Diego, California, Us

◦ Developed a toolbox that can automatically schedule the optimal quantization precision combination of different layers in a neural network.◦ Quantized popular deep learning models with the toolbox that achieved less than 2% of accuracy loss with aroughly 50% computational cost reduction.◦ Surveyed and summarized the most recent and popular neural network quantization and compression methods and schemes.

San Francisco, Ca, Us

◦Worked with a gaming analytics company, Mayhem, to build a robust monitoring stack.◦Deployed a clone of Mayhem’s server and the monitoring stack on to AWS with Terraform.◦Collected enhanced monitoring metrics from Mayhem’s RDSMysql cluster using Prometheus.◦Designed a dashboard withGrafanathat can accurately reflect resource allocation and usage of an RDS database as well as the error and slow query log flow.◦Created alarms for potential spike traffic and resource shortage for the database.◦Launched the whole system onto Mayhem’s real server system.

College Station, Tx, Us

College Station, Tx, Us

Designed course's coding homework and part of the final project.Guided the students for their final projects.Corrected and advised the students' coding homework

Guangzhou, Guangdong, Cn

◦Robot Operating System(ROS) based traffic simulator:∗a.: Developed a new feature that enabled external control from the outside agents using C++ and Python, used by the behavior planning teams in the company.∗b.: Built an API that allowed the ROS simulator log data with a particular frequency and dumped them in CSV files using python, used by our team for further training.◦Study on automatic ramp merging:∗a: Designed a new reward function for the reinforcement learning ramp merging that treats safety as apart of reward rather than a hard boundary.∗b: Built a DQN and DDPG agent-based on Keras using python, able to achieve a 0.24% collision rate with our reward function, comparable to current state-of-art method S-T graph.

Collected a huge amount (over 100000 pieces per day) of data from HUAWEI’s own git.Performed ​Apriori and ​Multi-level Based Data Mining Algorithm on judging if an engineer is facing a problem based on the frequency and the quantity of code the uploaded.Applied​ Spring MVC​ as the frame of the websiteManaged between different types of Databases using Java, including SQLServer and MySQL.

Doctor Of Philosophy - Phd, Electrical And Electronics Engineering

Master Of Science - Ms, Electrical And Electronics Engineering

Bachelor'S Degree, Electrical And Electronics Engineering