Skip to main content
EMEP’s quickstart is not a code walkthrough; the platform is documentation-first and implementation has not started. This page tells you how to navigate the specification set, what to read first for your role, and which quality gates the documentation must pass before engineering begins.

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

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.

Primary loop reminder

Continue to Context for the full project definition.