Nishad Gothoskar

I am a Research Engineer at Vicarious AI. Here, we are developing artificial intelligence for robots, drawing inspiration from the information processing systems in the human brain. By combining insights from neuroscience and probabilistic modeling, we seek to design algorithms that can match humans' ability to learn and generalize, unlike any AI system seen to date.

Previously, I was a Visiting Research Associate with the Learning and Intelligent Systems Group at MIT CSAIL advised by Leslie Pack Kaelbling and Tomas Lozano Perez. Here I worked on a variety of projects involving Task and Motion Planning, Bayesian Optimization, and RL.

Before this, I worked full-time as a Prediction engineer at Uber Advanced Technologies Group on autonomous vehicles. I built models for predicting vehicle interactions to better guide safe motion planning. I worked with Jeff Schneider and Ian Dewancker.

From 2014-2017 I was at the School of Computer Science at Carnegie Mellon University where I studied CS and Math and graduated in 3.5 years. My internships were are Google, where I worked with Scott Jenson on Physical Web, and zSpace Virtual Reality, where I worked on Stylus Gesture Recognition.

My research at CMU involved both Robotics and Applied Machine Learning. I worked on Bidirectional A* search based planning (Maxim Likhachev), Activity Recognition for Sensors, and ML for Mobile Privacy (Yuvraj Agarwal)

Email  /  Google Scholar  /  Github  /  LinkedIn

Learning, Task and Motion Planning System
MIT CSAIL, Learning and Intellgient Systems Lab, 2018

Built a TAMP system with Perception, Planning, Learning, and Execution modules. The perception was built using a Deep Learning object detector and a pointcloud based pose estimator. We demonstrated our system on a PR2 robot. Code will be published shortly.

Uber Advanced Technologies Group - Prediction Team
Uber ATG, 2017-2018

Developed a predictive model (as an intern) to resolve interactions between the autonomous vehicle and other road actors (vehicles, bikes, pedestrians). Then I deployed this model on Uber's self-driving fleet, measurably improving vehilce safety. Youngest full-time hire working alongside a team of PhDs.

SCS and Institute for Software Research @ CMU, 2017

The team developed an application to prevent privacy leaks on Android. Analyzing this I was able to identify that 3rd party libraries were the main causes of privacy leaks. To address this we proposed a new model of privacy protection for mobile devices. We proved the theoretical guarantees of such a model and after implementing it were able to reduce the amount of privacy sensitive data leaked by more than 25%. Was a Featured Paper at ACM IMWUT conference.
[paper] [CMU press release] [Cheddar Interview]

GIoTTO: Internet of Things
SCS @ CMU, 2016

The team is looking to build a general purpose sensing device that can be deployed across environments and be trained as synthetic sensors. The goal to is to build a robust machine learning layer that can be deployed to users for personalized usage. The specific research problem I am tackling is the issue of transfer learning. We want models to be able to transfer between spaces and sensors and maintain accuracy of prediction. This is particularly important in the IoT space where over huge deployments of sensors, it is infeasible to have models for each individual device.

Robotics Institute: Search Based Planning Lab
Robotics Institute @ CMU, 2017

I implemented a variety of Motion Planning algorithms (including variants of A*) for use and visualization using the Search Based Planning Framework. I tested implementations on randomly generated graphs as well as on a PR2 in simulation.


Uber Advanced Technologies Group - Prediction Intern

Google - Software Engineering Intern [github] - Top 5 Contributor

zSpace Virtual Reality - Research Intern [video] [paper]


William Lowell Putnam - Top 500

Fifth Year Scholars Program CS (Deferred Admission)

Dean's List All Semesters


Particle-based SLAM for Autonomous Vehicles





Fundamentals of Computer Science (15-112)
Principles of Imperative Computation (15-122)
Principles of Functional Programming (15-150)
Intro to Computer Systems (15-213)
Great Theoretical Ideas in Computer Science (15-251)
Green Computing (08-540)
Parallel and Sequential Data Structure and Algorithms (15-210)
Machine Learning (10-601)
Computational Programming and Problem Solving (15-295)
Probability and Computing (Honors) (15-359)
Algorithm Design and Analysis (15-451)
Parallel Computer Architecture and Programming (15-418)
Theoretical Cryptography (15-503)

source code