QuackTech Innovation
We build AI systems that learn how scientists evaluate research — turning years of tacit expertise into explicit, evolving judgment frameworks.
Capabilities
AI-driven systems that capture, formalize, and operationalize the evaluative frameworks scientists use to identify promising research directions.
Structured evaluation frameworks that capture a scientist's criteria for promising research directions, making tacit expertise explicit and reproducible.
Complete research loop from literature discovery and hypothesis generation to experimental planning — with taste models serving as directional filters at each stage.
Judgment standards that evolve through iterative human-AI collaboration — distinct from static reward models or community-level bibliometric signals.
Methodology
Our approach centers on a specific challenge: scientific taste is typically tacit, accumulated through years of domain immersion, and resistant to direct articulation.
We work closely with domain scientists to surface and formalize their evaluative criteria through structured collaboration.
Each round of feedback refines the system, producing increasingly accurate research direction assessments.
The evaluation framework itself improves with use, adapting as research priorities sharpen through discovery.
Impact
Team
Founder & Principal Scientist
Columbia PhD in Electrical Engineering. Building AI systems that bridge the gap between scientific intuition and computational methods.
Former Staff Applied Scientist at Samsara · Staff ML Engineer at Pinterest · Senior SDE at Microsoft
University Partners
Ongoing research collaborations with leading university labs in applying AI methodology to experimental sciences.
Technology & Strategy
Advisors from top technology companies providing guidance on product strategy, engineering infrastructure, and scaling AI systems.
Get In Touch