NAVIDBR Applied AI Systems
AI ideas are easy. Working systems are harder.
Business -> Systems -> Data -> AI -> Products
I work on applied AI systems where the difficult part is not the model alone. The work starts with the business workflow, source material, evaluation boundary, and evidence needed before an AI or product surface should be trusted.
navidbr.me is the canonical public layer for identity, notes, work records, and routing. GitHub holds source evidence for selected public proof records.
| Layer | Practical question |
|---|---|
| Workflow | What repeated decision, document flow, or handoff actually needs support? |
| Source preparation | What data, record, trace, or citation must exist before AI can help? |
| Evaluation | What behavior should be tested before a system is trusted? |
| Boundaries | What should be refused, escalated, or left to a human? |
| Product path | What is proven publicly, what is not proven, and what should happen next? |
Where a NAVIDBR work page exists, it states the claim and boundary while the repository provides inspectable source evidence. The standalone Evidence Loop project is an open-source reference implementation. These are public proof records, not client claims or production deployment claims.
| NAVIDBR work record | Source evidence | Public signal |
|---|---|---|
| Databricks CaseOps Lakehouse | Repository | Source preparation, provenance, validation, evaluation, and AI-ready handoff records. |
| AWS Bedrock CaseOps Control Tower | Repository | Grounded retrieval, citations, structured outputs, validation, and escalation boundaries. |
| Agent Behavior Evals Lab | Repository | Approval gates, refusal boundaries, uncertainty handling, traces, and quality gates. |
| E-commerce Purchase Intention MLOps | Repository | Reproducible ML workflow, evaluation, local serving, tests, and model-card notes. |
| Evidence Loop Visibility Engine | Repository and documentation | Bounded Loop Engineering for evidence-first SEO, AEO, GEO, and LLMO proposals without autonomous publication or outcome claims. |
- Public proof records:Evidence Loop, Agent Behavior Evals, Bedrock CaseOps, Databricks CaseOps, and Purchase Intention MLOps
- Evaluation and governance labs:Agent Behavior Evals, Agent Gate Demo, and OpenClaw Device Lab
- Workflow, data, and ML tools:Agentic Excel Review Template, StatAgent, and the public proof records above
- External projects: forks and upstream-derived repositories remain separate from NAVIDBR proof records
Business pressure
-> workflow and owner
-> source shape and constraints
-> evaluation and boundaries
-> AI or product surface
-> public proof before product claims
- Website: navidbr.me
- Email: navid@navidbr.me
- LinkedIn: linkedin.com/in/navid-broomandfar