Matt Shaffer
AI Research Scientist
& Co-Founder
matt@discovermatt.com
discovermatt.com
github.com/wx-b
github.com/planetceres
github.com/pylit-ai
Palo Alto, CA
Profile
AI research scientist and founder. Building and leading teams that turn emerging research into deployed, closed-loop systems.
Currently focused on agents that leverage tools, meta-reasoning, verifiers, evaluation-driven iteration, harnesses and human-AI collaboration that make frontier capabilities usable by real organizations. Leading technical efforts at RIOS toward high-leverage commercial applications.
Education
UC Berkeley
2018
Master of Information and Data Science
UC Santa Barbara
2003
B.A. Geography
Patents
US-11642796-B2
Tactile perception apparatus for robotic systems
US-11679418-B2
Automatically individually separating bulk objects
US-11433555-B2
Robotic Gripper with Integrated Tactile Sensor Arrays
US-11413760-B2
Flex-Rigid Sensor Array Structure for Robotic Sensors
US-11383390-B2
Robotic Workcell and Network
US-11273555-B2
Multimodal Sensor Array for Robotic Systems
Experience
RIOS Intelligent Machines
2018-PRESAI Research Scientist Fellow & Co-Founder
- • Raised $44M+ from seed to post Series B and scaled technical teams to deliver novel high-dimensional sensor modalities and deep learning control systems.
- • Deployed AI-powered robotics across multiple industrial domains, moving beyond POC demos to handle high-friction edge cases.
- • Repeatedly refocused technical efforts toward high-leverage commercial opportunities, ensuring frontier tech drove genuine ROI.
- • Led the transition toward agentic systems for industrial workflows, emphasizing evaluation gates and safety constraints.
Dishcraft Robotics
2018Deep Learning/AI Engineer Intern
- • Performed object recognition/detection and grasp prediction.
- • Implemented experimental frameworks to optimize CV models.
Qlabs.ai
2017Founder & Developer
- • Cloud-based assistant with NLP/SMS integration.
Meshrocket, Inc.
2016-2018Co-founder & U.S. Project Lead
- • Wireless-mesh technology for underserved areas.
Publications
"Pos3R: 6D Pose Estimation for Unseen Objects Made Easy"
2025A method for estimating the 6D pose of any object from a single RGB image, making extensive use of a 3D reconstruction foundation model and requiring no additional training.
"Differentiable Neural Surface Refinement (TNSR)"
2024Introduces transparent neural surface refinement explicitly incorporating physical refraction and reflection tracing.
"Ray Deformation Networks"
2024Deformable networks to learn tailored deformation fields for refractive objects in NeRF frameworks.