Machine Learning Engineer2022 - PRESENTLayer 6 (TD Bank)
Technologies: Java, Scala, Spark, Databricks, Azure
- Created data pipelines, productionization, and scaling ML models for TD, one of the largest banks in North America and the largest bank in Canada. Layer 6 is a leading research and applied AI branch of TD.
- Built development and production data pipelines on large customer datasets for various financial use cases using Spark, Databricks notebooks, and Azure.
- Involved in hands-on model development that includes feature engineering, model training, and inference.
Technical Co-founder2020 - 2021Anooka Health
- Developed a MERN-based web application from scratch. Conducted user interviews, designed a prototype, developed an initial MVP and final product, and managed two full-stack engineers and the development of the product.
- Performed in-depth research of 3D pose estimation and its applications in fitness, such as form feedback and rep counting, and designed the system that operates on the edge and in the cloud.
- Developed three MERN web apps—beta version for initial testing with users in five weeks, the final version as direct reports with two SWs and designer in three months, and a video-based partner exercising app in six weeks.
Machine Learning Engineer2019 - 2020Passenger AI
Technologies: Python, Deep Learning, Computer Vision, AWS, Machine Learning, Machine Learning Operations (MLOps), Kubernetes
- Built an online service to detect objects during cabin surveillance on AWS that was 30% more accurate yet as effective as the previous model for Passenger AI, a VC-backed startup revolutionizing safety monitoring for self-driving vehicles.
- Obtained incredible results (F1 score > 0.95) on action recognition that could run in real time on low-power NVIDIA Jetson Nano.
- Profiled and optimized concurrent on-device client code to efficiently execute business logic and neural network inference.
- Learned and experimented with computer vision topics like multi-view geometry, tracking, person re-identification, optical flow, and gaze estimation to understand how this could influence product in the short and medium-term.
- Researched optics, camera sensors, and lenses to understand how cameras could drive products. Proactively drove the transition to a new, more robust camera system that performed better across difficult imaging conditions, such as low light.
Machine Learning Engineer2018 - 2019Motion Metrics
Technologies: Python, C++, Computer Vision, TensorFlow, PyTorch, Machine Learning, Machine Learning Operations (MLOps)
- Developed a new deep learning architecture and real-time computer vision pipeline running on constrained edge devices that resulted in 3x improvements in business metrics and new features for the company's main cash cow product.
- Researched state-of-the-art methods for object detection, pose estimation, and action recognition and performed system design of full computer vision pipeline. One part of the contribution was published in the CVPR workshop.
- Contributed to the ML lifecycle, including research, prototyping and experimentation, data operations, evaluation, model optimization, and deployment.