VinayaBrhami
An operating system for a robot.
Three tiers with an explicit real-time boundary: an Operator on the robot, a Tactician arbitrating control, a Strategist planning off the critical path.
Built with
2026 · AI Systems
What it does
Three-tier architecture with an explicit real-time boundary: Operator → Tactician → Strategist
18 Rust crates, 154+ unit tests, zero-warning build
JAX training pipeline feeding Rust inference at the edge
Built on a reusable Rust robotics and agent crate ecosystem
The long version
Robotics stacks tend to grow as a collection of nodes that happen to talk to each other, with no agreed answer to which part is allowed to be slow. VinayaBrhami starts from that question instead: the architecture splits a robot into three tiers with an explicit timing contract between them.
L0 Operator owns the real-time edge loop — sensors, control, the things that must not miss a deadline. L1 Tactician turns intent into executable tasks and arbitrates control authority. L2 Strategist runs mission planning, the learning engine, and the memory ledger off the critical path, where taking a second to think costs nothing.
The runtime is Rust throughout, with JAX handling model training off-robot, and it sits on a reusable crate ecosystem meant to make agents, middleware, and real-time systems as approachable in Rust as they are in Python. Status is honest: v0.1, core architecture implemented and tested, not production-ready.