As part of the AI Foundations team, I am investigating applications of LLM and VLMs to generalize the Waymo driver.
I work on a range of projects including Pretraining/Torso-patching, and developing multimodal sensor encoders.
As part of the AI Foundations team, I am investigating applications of LLM and VLMs to generalize the Waymo driver.
I work on a range of projects including Pretraining/Torso-patching, and developing multimodal sensor encoders.
I lead the perception team working on the "Delivery Zip" - part of Zipline's next generation delivery platform that enables safe, precise and magical deliveries (check out this short video by Mark Rober!).
Built the entire perception stack from scratch, executing a multi-year roadmap from concepts to a production system serving tens of thousands of customers in multiple US metros.
Architected and delivered technologies spanning end-to-end vision based 3D occupancy, deep-learned stereo vision and semantic segmentation.
Previously, I led tracker development on the Acoustic Detect And Avoid program. Our system received FAA authorization for the first ever BVLOS deliveries in the US.
As a founding engineer at Aurora, I helped bring up a broad swath of the self driving stack in the early days.
I subsequently became a Technical Lead Manager on the perception team, leading teams on various object detection and tracking efforts. My day to day involved experiment design, ML model development, onboard integration, visualization and evaluation.
I also led perception bring-up on several platforms, from cars to trucks.
I helped found Jumpstart - Aurora's Returnship program to help bring women back into the workforce after a career gap.
I owned object tracking and evaluation pipelines for urban autonomous driving. Our work on vehicle tracking from lidar was awarded a US patent.