Python · FastAPI · LLM Applications · Bounded Agent Workflows · Production Reliability
I'm a production SQL Server DBA transitioning into applied AI engineering. My background includes incident response, performance troubleshooting, blocking and query analysis, HA/DR operations, backup and recovery, monitoring, automation, deployment support, and keeping production systems reliable.
I now apply that operating mindset to backend and AI systems built with Python, FastAPI, Pydantic, LLM APIs, structured outputs, tool calling, evaluation, Docker, PostgreSQL, REST APIs, and real production deployment.
| Area | Demonstrated technologies and practices |
|---|---|
| Core engineering | Python, FastAPI, Pydantic, SQL Server, PostgreSQL, Docker, Linux, Git, GitHub, GitHub Actions |
| AI applications | OpenAI Responses API, Structured Outputs, Tool Calling, Agent Workflows, Evaluation, Evidence Grounding, Tavily Search, REST APIs |
| Production & infrastructure | Nginx, Docker Compose, Ubuntu VPS, HTTPS, CI/CD, Observability, Production Debugging |
| Frontend | React, TypeScript |
|
A production-deployed controlled incident-response lab that investigates application and PostgreSQL failures through restricted diagnostics, produces evidence-backed findings, and requires human approval before allowlisted remediation.
Built with: Python · FastAPI · Pydantic · PostgreSQL · OpenAI Responses API · Docker · Nginx |
|
A production-deployed bounded research system that decomposes a question into 2–5 focused assignments, searches sources concurrently, preserves application-owned evidence, validates grounding relationships, and synthesizes a cited report.
Built with: Python · FastAPI · Pydantic · asyncio · Tavily · OpenAI Responses API · Docker · Nginx |
|
A production-deployed document extraction service that converts invoice PDFs and images into validated structured data and supports stateless questions over the extracted invoice.
Built with: Python · FastAPI · Pydantic · OpenAI Structured Outputs · pypdf · PyMuPDF · Pillow · Docker |
The model is only one component.
The engineering is in the system around it.
User / System
|
v
API Boundary
|
v
Application Workflow
|
+----+---------+
| |
v v
Model Tools / Data
| |
+------+-------+
|
v
Validation / Policy
|
v
Human / Application Control
|
v
Final Result / Action
Across all three projects, application code owns the boundaries: inputs, tools, IDs, validation, policy, failure handling, and what the model is allowed to influence.
|
|
|
I'm not leaving production engineering behind.
I'm applying it to AI systems.
→ What happens when the model is wrong?
→ What happens when a tool fails?
→ Can we trace where the answer came from?
→ Can the output be validated?
→ What actions require human approval?
→ How do we evaluate AI quality?
→ Can we reproduce a failure?
→ Can another engineer understand and operate the system?
|
|
|
RAG, embeddings, vector search, and MCP are learning areas—not features of the three systems above.
Production SQL Server DBA
│
▼
Production Systems & Reliability
│
▼
Python / Backend / Automation
│
▼
LLM Applications
│
▼
Bounded Agent Workflows
│
▼
Applied AI Engineering
Current learning priorities are listed separately above; no unfinished certification is presented as earned.
|
Live demos, architecture, project stories, and technical writing. |
Source code, tests, architecture, and engineering decisions. |
My path from production database engineering into applied AI. |
Building AI systems with a production engineer's mindset.
Python · FastAPI · LLM Applications · Bounded Agent Workflows · Production Systems