Service
Rust engineering
Production Rust — from CLI tools to async services to embedded targets.
- Typical engagement
- 8-16 weeks
- Capabilities
- axum + tower servicestokio + rayon for hybrid workloadssqlx-typed databasesasync lifecycle contractsFFI
We write Rust the way Rust wants to be written: typed-state machines, no panics in production, careful use of unsafe.
Much of our work starts here. New services, CLIs, library crates, FFI bridges — Rust is the substrate.
We treat the type system as the place where invariants live: typed-state machines for protocols, newtypes that parse rather than validate, Result over panic, small surfaces that can be reviewed in isolation. unsafe happens when it has to, and the safe wrapper around it is what your team consumes. Libraries expose honest async lifecycles — CancellationToken and TaskTracker, not a framework’s opinion.
Tests come first, consolidated and comprehensive. Logging is tracing with structured fields, never println; configuration is typed, never scattered .env. We build for the whole machine: tokio for concurrency — I/O, coordination, cancellation — and rayon for parallelism, saturating every core on the compute-heavy path, with a proper bridge between the two runtimes so neither starves the other.
This practice is codified, not aspirational: we maintain an in-house Rust 2024 toolkit for agentic development — 29 skills from ownership through Pin, kept in sync with Google’s comprehensive-rust, every pattern backed by runnable examples with 1,600+ embedded tests, warning-clean. Our AI agents write Rust under the same discipline we do.