Human-in-the-loop systems · Python · Data quality · Technical operations
I build practical AI and automation systems that turn operational workflows into explicit rules, inspectable outputs, and clear human handoffs. My technical-operations and data-quality background shapes how I approach reliability, integration, and failure handling.
These personal portfolio labs use synthetic data or sanitized operational evidence.
| Project | What it demonstrates |
|---|---|
| Support Operations Workflow Pack | Provider-neutral document retrieval, inspectable context assembly, a local model-command path, response guardrails, explicit human escalation, and a tested scorer for a bounded support assistant. |
| Data Integrity Validator | A Python/pandas CLI for normalization, validation, deduplication, referential-integrity checks, explainable rejection output, and automated regression coverage. |
| Azure Operations & Governance Lab | Reproducible Azure infrastructure, governance, observability, CI validation, teardown automation, incident-style documentation, and captured deployment evidence. |
- Context and harness engineering for reliable AI workflows
- Document-grounded assistants with business-specific knowledge and escalation rules
- Structured-output evaluation and citation guardrails for bounded assistants
- Developer tools that make automation easier to inspect, operate, and trust
- Model the real workflow and its failure modes.
- Automate the repeatable work.
- Keep consequential judgment behind explicit review or approval.
- Emit evidence another person can inspect and reproduce.