Read paths by role
Research Engineer
Start with the Research Evidence Map, then merge math and the Evolution Engine spec.
ML Engineer
Start with Model Compatibility, Merge Engine, and Tensor Operations.
MLOps Engineer
Start with GPU Orchestration, Storage Architecture, and Offline Deployment.
Backend Engineer
Start with API Specification, CLI Specification, and Configuration.
Evaluator
Start with the Evaluation Framework, Benchmark Catalog, and Reproducibility Protocol.
Project Manager
Start with the Roadmap, Phases, and Tasks.
Documentation quality gate
Implementation must not begin until every item below is verified. Each item is enforced by the Consistency Report and Traceability Matrix.1
Every major architecture section has a diagram
See the Diagram Index.
2
Every important algorithm has a flowchart
See the Algorithm Flowchart Index.
3
Every lifecycle has a state machine
Model, experiment, candidate, and artifact lifecycles are all explicit.
4
Every critical interaction has a sequence diagram
Import, compatibility, merge, evaluation, evolution, deployment, rollback.
5
Every requirement maps to a task and a test
Enforced by the Traceability Matrix.
Evidence classification
Every claim in the documentation is tagged as one of:
- Established research: peer-reviewed and reproduced.
- Research hypothesis: plausible but not confirmed in EMEP’s setting.
- Engineering assumption: chosen for design, not empirically validated.
- Experimental result: recorded with an Experiment ID.
- Future goal: planned but not implemented.
- Unsupported claim: flagged and must be removed or downgraded before publication.