Rust ML LLM Architecture

Separate orchestration, ML runtime, data, API, evals, and deployment concerns; treat model output and retrieved documents as untrusted inputs.

Process

1. Identify access patterns, trust boundaries, runtime constraints, ownership, and operational requirements before choosing structure.
2. Prefer the simplest design that satisfies current requirements and leaves clear extension points for known near-term changes.
3. Make data flow, failure handling, observability, and rollout constraints explicit.
4. Translate the design into concrete implementation and review checks before coding.

Coordination

Use this skill together with itsol-task-intake for ambiguous work, itsol-self-review before handoff, and focused security-* or infra-* skills when the change touches trust boundaries or deployment behavior.

Focused References