Developed tracking and sensor fusion algorithms in a team of 6.• Reduced flickering orientation estimates.• Improved data association to reduce track drops.• Improved velocity and acceleration estimates.• Reduced track drift by implementing a better measurement to track alignment algorithm.• Implemented algorithm to track objects of arbitrary shapes in a team of 5.• Optimized lidar scan pattern for better short- and long-range point density.• Added tracking visualizations for object velocity, covariance, and shape.

• Focused on object tracking and radar-related machine learning.• Incorporated dynamic object motion models into labeling pipeline to improve both label throughput and quality.• Implemented ML model for radar-only point segmentation and improved a multi-sensor object detection model by adding radar.• Transitioned a learned model for detection-to-track association from Tensorflow to PyTorch.

• Developed computer vision algorithms to detect and track moving objects using feature detection, feature matching, optical flow, and image alignment using a homography and the essential matrix.• Implemented Kalman filters and probabilistic data association to associate detections with state estimates. • Optimized computer vision and tracking algorithms to run in real-time on an embedded processor.

Improved internal company webpage. Optimized SQL queries, added graphs, scheduled data migration, and Integrated employee profile page with Google Calendar.

Programmed 10 drivers in C to control third-party lighting, security, camera, and shade systems from an embedded microprocessor using HTTP, TCP, and RS-232 protocol.