Lead a team of two developers in integrating an advanced anomaly detection model into the Aurora Imaging Library (AIL) for training and inference in C++. Conducted research and experimentation on multiple state-of-the-art anomaly detection models using PyTorch in Python. Designed and developed AIL's deep learning image segmentation and object detection training pipelines in C++ using MXNet. Accelerated inference using OpenVINO and CUDA execution providers for ONNX Runtime. Streamlined the data augmentation process with a customizable automated pipeline. Implemented a user-friendly system for creating, managing, and storing segmentation and object detection datasets, reducing memory usage by up to 5x. Guided and trained junior developers and interns to develop their skills and maximize impact.

Senior Computer Vision Software Developer at Zebra Technologies

July 1, 2024 - Present

Led end-to-end development and training of an object detection system for hazardous materials labels, from proof-of-concept to deployable C++ inference library. Designed and delivered technical demos and proofs of concepts to engineering leadership, product managers, and other stakeholders, driving alignment on feasibility and scope. Collaborated with product and project managers to define roadmaps, timelines, technical risks, and deployment constraints for production integration. Led development of a state-of-the-art OCR model, improving text recognition accuracy and speed. Migrated deep learning training algorithms from MXNet (C++) to PyTorch (Python), improving efficiency and maintainability. Collaborated with UI/UX developers, sales engineers and customer support to enhance customer experience. Conducted rigorous code reviews for four developers, ensuring high-quality, maintainable code.