Experience

Head of Sim Realism, Director of Engineering

(Jan 2021 - Present)

Leading Sim Realism at Waymo, an org of over 100 engineers and research scientists across multiple locations in US and UK. Responsible for ensuring the realism of all simulations at Waymo, both at the raw sensor level (cameras, lidars, radars etc.) as well as the sematic level (bounding boxes, perception objects and attributes etc.). The simulated world includes intelligent realistic agents (vehicles, cyclists, pedestrians, animals etc.) as well as other dynamics such as traffic lights, weather, lighting, traffic conditions etc. 

This involves building massive foundation world models, fine-tuning, distillation, integrations in to downstream workflows and evaluation of the models and the E2E systems. Extensive use of and research in SOTA ML methods including self-supervised auto-regressive FMs, diffusion, 3DGS, NeRF, parameter efficient fine tuning, multi-agent RL, multi-task adaptation etc. TPU based fully differentiable simulator to house these models for rapid evaluation and closed-loop training of the Waymo driver. Synthetic data generation for data augmentation as well as model evaluation. 

Previously, built a 35 person Behavior Models team that improves the quality of predicted behavior of agents that are of interest to the Waymo Driver. This included generating high quality inputs in appropriate representations to Deep Learning models, enforcing real world physics and other constraints to ensure predicted trajectories are realistic, monitoring for anomalies and failures, both real time and in simulations, and fixing any issues detected and performance evaluation of various models and systems using advanced data science methods.

Head of Google+ Ranking and Recommendations, Staff Software Engineer  (Oct 2012 - Jul 2017)

Tech Lead and Manager of Google+ Ranking and Recommendations team responsible for stream ranking (home, communities, notification, topics, explore, hashtag streams), recommendations (suggested posts, communities, collections, topics), search (for posts, collections, people, communities, etc.).

Joined the team as an L4 SWE in Q4 2012, promoted to Senior SWE in Q4 2014 and to Staff SWE in Q2 2016. Became the TLM of the team in Q3 2015 and grew a small team of 8 engineers (2 Senior SWEs, 4 L4 SWEs, 2 L3 SWEs) to large team of 25 engineers (7 Senior SWEs, 5 L4 SWEs, 13 L3 SWEs). 

Moved the core metrics by double digit percentages via multiple innovations evidenced by 10 patents. Simplified and modernized the infrastructure to clear tech debt and to enable incorporation of state of the art ML technologies. Expanded team’s ML technologies from purely regressions to DNNs, PCA, HAC, MAB, MDP etc.

Kept the morale high through multiple re-orgs and product changes as evidenced by 0% attrition and high approval ratings: 95% (2017), 93% (2016) favorable over all questions in Google Manager Feedback Surveys (MFS). Built a culture of high performance as evidenced by 100% team promo success rates in 11 promotion attempts (6 L4, 4 L5 and 1 L6 promo) based on multiple high impact launches.

Duke University, Research and Teaching Assistant (Aug 2007 - July 2012)

Research into tiling theory, algorithmic self-assembly, complexity theory, probabilistic analysis, structural DNA nanotechnology culminating into many publications. Taught courses in algorithms and data structures as a Teaching Assistant to Prof. John Reif.