SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Transfyr is building physical AI systems for scientific research. The company captures real-world laboratory execution data and transforms it into machine-readable records that help scientists learn from failures, transfer knowledge, train the next generation, and provide grounded data for AI and robotics. Backed by a $25M seed round with advisors including Chris Ré (Stanford), David Baker (Nobel laureate), and Kevin Weil (former CPO at OpenAI), Transfyr is tackling frontier problems at the intersection of science, perception, machine learning, and robotics.
As an AI/ML Engineer, you will build end-to-end machine learning systems that learn from messy, real-world data captured in active laboratory environments. You'll work closely with computer vision and software engineering teams to integrate grounded data (object coordinates, action timings, etc.) and develop interpretations of scientific actions and their impact. Your models must handle partial observability, significant noise, long context requirements, changing protocols, and ambiguous outcomes while producing trustworthy signals for scientists and downstream systems.
Key responsibilities include: designing multimodal learning pipelines that combine vision, audio, sensor data, and metadata; developing models that explain why experiments succeed or fail; building systems that surface model confidence and uncertainty; integrating models into real workflows where outputs influence human decisions and robotic actions; ensuring models learn transferable structure across labs and geographies; and laying groundwork for future physical AI automation.
You will demonstrate high agency, bias toward action, and comfort with ambiguity. You identify what needs to be learned, build appropriate scaffolding, prototype quickly, test assumptions against real data, and iterate based on failure. You understand when sophistication helps versus obscures, and you can make progress when labels are incomplete, feedback is delayed, and success criteria evolve. Strong ML fundamentals, solid software engineering judgment, and the ability to work collaboratively across disciplines are essential.
