Engineering Requirements Copilot
Outcome
Keep automotive systems requirements traceable and testable by linking each requirement to design decisions, verification cases, and applicable regulations, with engineering review before a baseline is frozen.
| Portfolio attribute | Value |
|---|---|
| Primary persona | Systems engineer |
| Primary aggregate | RequirementBaseline |
| Cosmos DB containers | 2 |
| Grounding data | requirements, designs, tests, regulations |
| Development estimate | USD 60-220 monthly dev estimate |
| Runtime options | Python FastAPI and C# ASP.NET Core |
Productivity hypothesis
These measures reflect the workflow, evidence, decision boundary, and industry risk for this use case.
| Impact type | Measure | Conservative internal assumption | Customer measurement |
|---|---|---|---|
| Productivity | Requirements review cycle time | 8-15% reduction | (baseline median review minutes - assisted median review minutes) / baseline median review minutes * 100 |
| Quality or risk | First-pass acceptance | 5-10 percentage-point improvement | assisted percentage accepted without material rework - baseline percentage accepted without material rework |
Important: Conservative internal pilot assumptions only; not public benchmarks, commitments, or expected customer results. Validate against a comparable baseline or matched control cohort with mandatory human review.
Rationale: Systems engineers can accelerate traceability analysis while keeping change-sensitive engineering decisions under human control.
Pilot guardrail: Maintain strict traceability across requirements, designs, tests, and regulations with formal engineering review.
Customer validation plan
- Baseline the current workflow with representative users and tasks.
- Run the accelerator on the same task mix with human review retained.
- Measure requirements review cycle time and first-pass acceptance with the documented formulas.
- Replace internal assumptions with customer-observed results before a business case.
Why Azure Cosmos DB
This AI solution needs an operational data plane, not only a vector index:
- Hybrid vector and full-text retrieval combines semantic recall with exact identifiers, codes, names, and domain language.
- Hierarchical partition keys and metadata predicates keep tenant and domain boundaries in the normal query path.
- Operational JSON, source chunks, workflow and agent state, feedback, and audit records remain together for low-latency reads and governed writes.
- Change feed reacts to source and operational updates without polling.
- Autoscale throughput, configurable consistency, and multi-region distribution support variable demand and enterprise availability.
Domain workflow

| Stage | Actor | Input → output | Approval |
|---|---|---|---|
| 1. Author or import the requirement | Systems engineer | Requirement text with allocation and verification method → SystemRequirement in the working baseline | Policy controlled |
| 2. Check quality, ambiguity, and duplication | Requirement quality agent | Requirement text and existing baseline → Quality findings and near-duplicate candidates | Policy controlled |
| 3. Link regulations and design decisions | Traceability agent | Requirement, regulation clauses, decision records → Bidirectional trace links with confidence | Policy controlled |
| 4. Assess verification coverage | Verification analytics service | Requirements and verification cases → Coverage report with unverified requirements | Policy controlled |
| 5. Freeze the baseline | Chief engineer and systems authority | Coverage report and open change requests → Frozen baseline released to development partners | Human required |
Technical architecture

Cosmos DB data model
| Container | Hierarchical partition key | Documents | Search paths |
|---|---|---|---|
requirementBaselines | /tenantId, /vehicleProgramId | RequirementBaseline, SystemRequirement, DesignDecisionRecord, VerificationCase | /requirementVectorEmbedding, /rationaleVectorEmbedding, /requirementVector, /rationaleVector |
homologationRegulationLibrary | /tenantId, /marketCode | RegulatoryRequirementClause | /clauseVectorEmbedding, /clauseVector |
Query contract
| Query | Container | Scope | Purpose |
|---|---|---|---|
| Requirement trace assembly | requirementBaselines | single-partition | Read a requirement with its decisions and verification cases in a single program partition query. |
| Unverified safety requirements | requirementBaselines | single-partition | Find high integrity level requirements without a passing verification case. |
| Market homologation gap | homologationRegulationLibrary | single-partition | Knowledge-container query against homologationRegulationLibrary (partition-aligned to /tenantId + /marketCode): list regulation clauses in force for a target market, then diff them against the programme's linked requirements; the library stores one clause document per market so the market read is partition-aligned. |
| Failing verification runs | requirementBaselines | single-partition | Surface verification cases that failed since the last baseline review. |
Python and C# services route each document type to its declared container and hierarchical partition key. ETag-based writes protect mutable aggregates, while append-only records and bounded cross-partition queries preserve scale.
Azure deployment

Clone the source repository, authenticate with Azure, and run:
Set-Location accelerators/automotive-mobility/engineering-requirements-copilot
az login
./Deploy.ps1 -Runtime python -Location eastus2
Change python to csharp to deploy the .NET implementation.
Explore the implementation
- Launch signed-in demo
- Implementation article
- Domain data model
- Workflow and API contract
- Python application
- C# application
- Domain container Bicep
- Platform Bicep
- Cost and configuration
- Source repository