Research Assistant, University of North Carolina at Charlotte, USA
- Developed a distributed messaging framework (in Golang) for computer vision applications at the Edge that achieved 10.1x end-to-end latency improvement over the state of the art messaging systems.
- Developed a low latency in-memory distributed storage (in Golang) uniquely tailored for computer vision applications deployed on resource constraint IoT embedded nodes.
- Designed a WiFi latency controller (in Python) for an IoT video surveillance system deployed at the Edge that can deliver real time video frames to latency-sensitive object/event detection applications within a settling time less than 1.5 second.
- Built an Edge test bed for latency measurement (less than 15 microsecond accuracy)
- Identified optimal keyframe selection policy for Edge video workloads that achieved 63.1% more efficiency
- Determined the best scalable Edge computing infrastructure (with 64 nodes) for real-time video analytics using NS3 simulation framework (in C++)
- Identified 630 unique image modifications in OpenCV that result up to 90% more object detection accuracy