Chukwuma Akpu

I’m a Senior Computer Vision Engineer with 6+ years building real-time perception systems from research through production deployment. My work focuses on object detection, multi-object tracking, and motion analysis optimised for edge hardware, with a strong emphasis on end-to-end pipeline architecture.

I’ve published first-author work at ICPR 2024 on a novel detection-and-tracking pipeline (YOLOv8 + SAM + DeepSORT), and I’m recognised for translating complex CV research into production systems that perform under real-world constraints. I’ve shipped high-performance models to resource-constrained platforms with measurable impact, including 97% MOTA tracking accuracy and major improvements to inference latency and labeling efficiency.

In my current role, I architected and deployed a real-time detection-and-tracking pipeline that handles occlusion, re-identification, and variable lighting, achieving 97% MOTA. I reduced inference latency by 71% using TensorRT and ONNX Runtime optimisation with FP16/INT8 quantisation, enabling sub-100ms perception loops on NVIDIA Jetson and other embedded GPU-constrained platforms.

I also lead generative vision research for synthetic dataset generation using LoRA and IP-Adapter, solving data scarcity when collecting real training data is expensive or infeasible. Beyond core computer vision, I bring expertise in SLAM, 3D reconstruction, sensor-driven systems, and I mentor engineering teams while maintaining CI/CD infrastructure for reproducible model training, evaluation, and deployment.