SSL Library Tracing:• Explored the openSSL libraries for Go and C++ and identified the best point in each library to attach an eBPF (a Linux tracing tool) UProbe (User Space Probe) for network data tracing.• Implemented SSL tracing with Uprobes to collect unencrypted data just received or about to be sent across network, which would otherwise be encrypted if traced through syscalls.• Led the development of this core feature in a team of three engineers, and integerated with existing… Show more SSL Library Tracing:• Explored the openSSL libraries for Go and C++ and identified the best point in each library to attach an eBPF (a Linux tracing tool) UProbe (User Space Probe) for network data tracing.• Implemented SSL tracing with Uprobes to collect unencrypted data just received or about to be sent across network, which would otherwise be encrypted if traced through syscalls.• Led the development of this core feature in a team of three engineers, and integerated with existing data collection pipeline.Automated Testing Framework:• Simulated real network traffic for a MySQL server, and implemented an automated testing pipeline to generate test cases from data captured by WireShark. Show less

Domain-aware Dynamic Networks• Proposed a novel domain-dependent Mixture-of-Expert(MoE) model for object detection and semantic segmentation in autonomous driving scenarios.• Implemented the model based on MaskRCNN using Pytorch, and conducted experiments.• The model achieved 2.6% higher AP50(Average Precision) than a static network on the BDD100K object-detection benchmark with no increase in inference-time computation.

MySQL Data Collection/Parsing Pipeline:• Implemented eBPF KProbes (Kernel Space Probe) that trace syscalls in Linux kernels to collect MySQL traffic.• Researched MySQL client/server protocol and implemented the detecting, parsing, stitching, and pairing process for MySQL request and response packets. Integrated with the existing data processing pipeline.• Tested and deployed the MySQL data collection, parsing and rendering modules on real Kubernetes applications.

Scalabel, an open-sourced web tool for efficient labelling of computer vision datasets:• Built a gateway server in Go that manages object detection models, allocates GPU resources to these models, and facilitates communication between the frontend web server and the backend model servers.• Used gRPC to pass the machine generated labels to the web server, and the human-finetuned labels back.• Benchmarked the network and computational latency of different object detection models.

Develop computer vision algorithms to analyze big medical data

•Contributed to a research publication that determines the friction factors of two phase pressure drop through a porous medium•Used LabVIEW software and a sophisticated experimental rig to derive conclusions on the validity of previous literature•Produced successful and efficient results by improving a previously unsuccessful experimental setup