End to end ML pipelineDeveloped wound segmentation model achieving 200% accuracy boost and 500% performance boost

• Developed and optimized machine learning models for production-level code, utilizing Python and TensorFlow to achieve scalability while conducting hyperparameter tuning and added respected object detection and classification metrics.• Worked Java/Spring Boot-based backend workflows to support efficient machine learning model deployment and integration.

• Worked on deep learning based automatic traffic counting system.• Added JS based vehicle tracking module.• Setup ML pipeline to intake raw video, process it and outputs count in csv formats.

•Lead efforts to detect financial fraud using machine learning and deep learning methods to develop models that can successfully recognize and stop fraudulent activity.•Created massive finance datasets from scratch using approaches for data cleaning, modification, and restructuring. In-depth analysis and visualization were then used to find trends and guide thoughtful, confident judgments.

• Provided computer vision services to a start-up in US, worked on development of automated checkout shopping experience, in direct competition with Amazon Go. • Work on development of recognition, detection and depth modules for computer vision pipeline. • Work on integeration of deeplearning modules with hardware sensors. • Worked on manuplation of terabytes of data. • Built autmoted annotation tool for the computer vision processes. • Working on table recognition tool using deeplearning and heuristic approches.

• Taught Deep Learning and Computer Vision from beginner to advance level.• Classes were amalgamation of lectures and practical assignments.• Pupil consisted of students and professionals from different backgrounds.

• Worked on real-time detection and tracking of different species of fish underwater underunconstrained environments.• Implemented Object detection neural network such as YOLO v2 on conventional videos andultrasound imagery to localize and track school of fish.• Worked on my senior year project, that was standalone device for vehicle surveillance and smart traffic management.• Used TensorRT along with other optimization techniques on state-of- the-art Objectdetection neural network,• Implemented neural network of reconfigured Nvidia Jetson TX2.• It provided real-time car detection, car identification and License plate localization.• Created a custom dataset of around 20k Pakistani number plates through synthesizing andaugmentation.