Trajectory Prediction for Cruise Autonomous Cars
Cruise Origin
- Developed data, pre-processing and modeling pipeline for Transformers, GNNs-based trajectory prediction architecture that acts on vectorized map and agent information.
- Improved prediction quality through lossless and long-range feature representation, reduced prediction latency by 20%, and improved training and validation speeds by 33%.
Benedict Florance Arockiaraj
ML Engineer
My research interests are at the juncture of deep learning and computer vision.