Looking for a job

Shaurya Mantrala

@shauryamantrala

AI/ML engineer at Carnegie Mellon University developing zero-shot robot training and a dataset of 1M manipulation trajectories.

United States

Message

At Carnegie Mellon University, I developed zero-shot techniques for robot training in wet labs and built a dataset of 1M object manipulation trajectories. I’m finalizing a Nature-targeted paper with Dr. Jose Lugo-Martinez.

At NASA Jet Propulsion Laboratory, I built the Atlas pipeline, using a graph-based knowledge system for robot training that achieved 80%+ policy accuracy across MSL archives. I also co-authored LUMIN with a JPL Lead Data Scientist.

As Chief Product Officer at SoberRide, I architected a quantum-optimized intoxication detection system. I presented as principal author on IntoxDetect V2 and QIntox, and SoberRide exhibited at CES and received recognition in the Conrad Challenge and Gore Innovation Excellence.

Experience

Work history, roles, and key accomplishments

Current

Lead ML Engineer

Current

Carnegie Mellon University

Oct 2024 - Present (2 years 1 month)

Developed novel zero-shot techniques for robot training in wet labs; built dataset of 1M object manipulation trajectories. Finalizing Nature-targeted paper with Dr. Jose Lugo-Martinez.

Zero Shot Robotics Training Wet Lab Object Manipulation Dataset Creation Nature Paper

Founding Engineer

Flywheel

Jun 2026 - Aug 2026 (2 months)

1M ARR bootstrapped startup; building marketing engineers. Developed E2E ICP-based content generation pipeline and testing eval. gates with 95% accuracy for enterprise MCP server.

ICP Content Development Evaluation Gates MCP Servers Enterprise Bootstrapped Startups

Robotics Engineer

Stealth Company

May 2026 - Aug 2026 (3 months)

Built SOTA tele-op gloves for data collection along with pipelines for MoCap datasets to train RL policies. Developed world state engine for actions in zero-shot contexts; tested on ASIMOV robot. 50M Seed Round; backed by Fei-Fei Li, 8VC, and more.

Hand Tracking Gloves Data Collection Motion Capture Data Collection Reinforcement Learning Policies World State Engine Zero Shot Contexts Asimov

AI/ML Researcher

NASA Jet Propulsion Laboratory

Jun 2025 - May 2026 (11 months)

Built Atlas pipeline with graph-based knowledge system for robot training, achieving 80%+ policy accuracy across MSL archives. Co-authored LUMIN with JPL Lead Data Scientist Dr. Emily Dunkel.

Graph Based Systems Robotics Training Policy Accuracy MSL Archives LUMIN Hybrid RAG CLIP Based KG Search

Chief Product Officer

SoberRide

Oct 2024 - Feb 2026 (1 year 4 months)

Architected triple-modular quantum-optimized intoxication detection. First high-school exhibitors at CES; Conrad Challenge Power Pitch (5/2,000) and Gore Innovation Excellence (2/1,200). Presented at MIT as principal author on IntoxDetect V2 (83% on 10K images) and QIntox (99.74%).

Quantum Optimized Detection Intoxication Detection IntoxDetect V2 QIntox Image Classification Presentation Skills