CTO of Bifrost

CTO & Co-Founder at Bifrost

RoboticsPhysical AIComputer VisionGenerative Neural NetworksSimulation InfrastructureDeep Learning

About

I'm the CTO and Co-Founder of Bifrost, where we build the simulation infrastructure necessary for the next generation of robotics and physical AI. My journey started in AI research, where I developed models for clinical applications—including one that outperformed Google's AI in trials—before moving into the world of simulation. At Bifrost, we're backed by Sequoia and Lux Capital, and our tech is used by the likes of NASA and Anduril to bridge the gap between lab performance and real-world reliability. I'm deeply passionate about solving the 'perception bottleneck' and helping teams iterate faster through robust data engines. I'm currently looking to connect with sharp technical talent, specifically Product Managers, and fellow researchers who are pushing the boundaries of how machines interact with the physical world.

Networking

What I can offer

›Solving perception bottlenecks for robotics teams
›Improving reliability in production through simulation
›Insights into synthetic data generation

Looking for

›Product Manager with sharp technical instincts
›Researchers and engineers in the robotics and physical AI space
›Expanding my professional network

Best fit for

Robotics researchersAI engineersTechnical Product ManagersFounders in the physical AI space

Current Interests

Physical AIDigital twinsPerception bottlenecksLong tail edge casesLab-to-production transitions

Background

Career

Transitioned from AI research at Medios Technologies to co-founding Bifrost, a simulation infrastructure company for robotics.

Education

Singapore University of Technology and Design (SUTD), focus on Robotics.

Achievements

›Co-founded company backed by Sequoia, Lux Capital, and Airbus Ventures
›Shipped an AI model that outperformed Google’s AI in clinical trials
›Built AI model for detecting potential blindness using iPhone cameras
›Infrastructure used by NASA, Anduril, and NTT Data

Opinions

Iteration speed is the primary competitive advantage for physical AI.
High lab scores are misleading; evaluation pipelines often fail to cover 70% of real-world failure scenarios.
Simulation should start long before physical hardware is deployed.
Automation should handle dangerous or dull work so humans can focus on what actually matters.