Led by a deep learning researcher, we are building Agentic Fellows, an autonomous research intelligence system.
Scientific breakthroughs rarely emerge from within a single field. The solution to a hard research problem often already exists elsewhere, described in a different vocabulary by people who don’t know you exist. Today, finding these connections requires either exceptional breadth or luck. Research teams spend months pursuing directions that experts in adjacent fields could have redirected in an afternoon.
Agentic Fellows closes this gap. Given a research problem, the system surfaces mathematically grounded solutions from other domains with verified structural equivalences, paired with a translation guide that researchers can act on immediately. It is verification-first by design: nothing reaches a researcher until the system has attempted to break it. Every output is fully traceable to its source. The system runs entirely within an organization’s infrastructure – research never leaves its walls.
The MVP is currently in progress.
