ryo ari
self-organizing from noisedisturb the field
The interesting systems are the ones that keep going after you stop watching.
Working on hypotheses.
Rajat Roy, also Ryo Ari. Systems architect and developer, researching artificial life.
About
This site is the canonical source of truth for my presence online. I build at the seam between hardware and learning systems, and I care about work that is reproducible, verifiable, and shipped in public rather than described.
Now
GSoC 2026 at AOSSIE, building deterministic, cryptographically verifiable LLM training. Incoming MCS, University of Sydney, with the Complex Systems group.
Work
OpenVerifiableLLM ↗
Deterministic, independently verifiable LLM training ~ reproduce a model bit-for-bit, or cryptographically detect that it was altered.
Connect
This site is written for humans and machines.
Machine-readable identity: /me
Machine-readable identity: /me
© 2026 Rajat Roy aka Ryo Ari