Experience

Oct 2021 - Present Senior ML R&D Engineer at Bodymapp Ltd

Bodymapp introduces the concept of body scanning to consumers by leveraging Computer Vision, Image Processing, Machine Learning and depth sensing technology in mobile devices. Bodymapp renders a 3D avatar of the user’s body, providing body measurements for health and fitness tracking.

• Led R&D sub-division using agile and scrum workflow to develop reconstruction part of the Bodymapp back-end
• Organized and led daily R&D stand-ups and weekly meetings with C-Level executives, briefing on progress
• Researched and engineered utilities for unstructured data processing using machine learning and computer vision libraries such as scikit-learn, pandas, PyTorch, TensorFlow, OpenCV, and Open3d
• Devised supervised training pipeline to undistort the data addressing the acute problem of distorted iPhone depth maps
• Developed in-house pre-release Systematic Evaluation metrics to enable data-driven decisions for measuring the accuracy, robustness and aesthetics of back-end output using ML concepts
• Designed a robust processing pipeline for calibration, cleaning, rectification and conversion of noisy in-coming data from iPhone sensors to prepare them for the registration and reconstruction stage
• Devised the complete post-processing module to create an appealing visualization of the outcome
• Leveraged cloud pipeline to perform rapid large-scale accuracy and aesthetic evaluations of releases
• Followed the complete cycle of the CI/CD pipeline according to the changing needs
• Used AWS DevOps and MLOps pipelines (bash, python, s3, ec2, cloud watch) for development and deployment and batch processing
• Displayed a dedication to ongoing learning and remaining at the forefront of technological advancements by keeping abreast of the latest research in the field

Dec 2019 - Oct 2021 Senior ML Engineer at NSW Department of Primary Industries-UTS

The project was conducted by the NSW Department of Primary Industries in collaboration with UTS worth $1.2M, aiming to automate cattle scanning for trait estimation by eliminating manual ultrasound scanning. By capturing 3D data via an advanced sensors set-up and processing via Computer Vision and Image Processing, and Machine learning the system can estimate fatness and muscle scores to increase productivity for the beef industry.

• Developed computer vision and image processing algorithms to capture, convert, pre-process and clean the data achieved via 3D sensor setup
• Devised a novel approach to reconstruct a model of a deformable object leveraging machine learning and computer vision libraries
• Developed and programmed two types of intrinsic and extrinsic online calibration methods for a rig of depth sensors as part of robotics perception
• Implemented my proposed non-rigid registration framework
• Conducted and programmed comprehensive error analysis and evaluation using ML models on the devised non-rigid registration and reconstruction
• Simulated the cattle scanning setup with depth sensors on ROS and Gazebo Platform
• Attended periodic meetings and pitched the progress of the project to the stakeholders
• Carried out field trials for testing the setup and pipeline, as well as collecting the data
• Implemented and evaluated the state-of-the-art methods for scanning
• Mentored and advised junior engineers and students

July 2015 - Dec 2019 PhD ML Research Scholar at UTS Robotics Institute (UTS:RI)

UTS Robotics Institute UTS specializes in delivering customized robotics and software solutions to maximize productivity, improve quality and safety, and generate efficiencies for commercial and government partners. The reconstruction of deformable objects project was defined as part of an MLA (Meat-Livestock Australia) funded research project aiming at detecting the best position of sensors to acquire rich data and then registering non-rigid frame streams and reconstructing them.

• Analyzed and processed data of cattle using state-of-the-art methods in ML and CV to register non-rigid data
• Created an optimization framework to find the depth sensor position and orientation for best view scanning
• Conducted field trials in Armidale to capture and collect data via our manufactured rig consisting of 16 depth camera
• Reviewed and simulated current methods of ML/AI/CV for scanning and object reconstruction
• Taught various fields, including C++ robotics etc
• Contributed to ML/CV and robotics community by Publishing papers in prestigious conferences and journals in the field of Robotics and ML
• Co-supervised master students on their thesis

July 2013 - July 2015 ML Engineer at Sahand CVIsion Lab

Sahand CVision Lab, associated with Sahand University of Technology, serves as a focal point for various stakeholders to identify and fulfill their technological requirements in a cost-effective manner. This is
achieved by engaging and leveraging the skills of young engineering professionals and researchers.

• Devised hand gesture recognition system as a virtual mouse to facilitate human-computer interactions
• Developed video processing method using ML/CV libraries to detect and track the fingertips
• Created ML model to translate and map the trajectories to mouse functions
• Applied ML and pattern recognition methods on an optical character recognition project for car tracking and plate detection and recognition in the Sahand campus