MARVIN · AI ENGINEERING
I build reliable AI systems, shaped by years of production database work.
I started in production database operations, where reliability, incident response, automation, and failure handling mattered every day. Now I bring that same mindset to AI systems built with Python, FastAPI, LLMs, and agent workflows.
- Production systems experience
- Python / FastAPI
- Live AI projects
01 · PROJECTS
AI systems I built to solve real engineering problems.
These projects are public, production-minded versions of the kinds of AI systems I want to build professionally—systems with clear workflows, strong validation, observable behavior, and deliberate failure handling.
Incident Investigation Agent
I built a real working AI system that creates safe, controlled failures in an application or database, investigates them using only approved tools, and does not fix anything automatically without human approval.
- Step-by-step investigation
- Conclusions backed by real evidence
- Only approved fixes, with human approval
- 01Controlled incident
- 02Restricted diagnostics
- 03Evidence-backed report
- 04Human approval
- 05Recovery verification
- Root cause
- Validated evidence
- Audited recovery
Research Agent
I built this tool to take a broad question, break it into smaller research tasks, search multiple sources at the same time, and produce one clear report with evidence and citations.
- Breaks questions into 2–5 research tasks
- Researches multiple sources in parallel
- Tracks evidence behind every claim
- Executive summary
- Cited findings
- Conflicts + uncertainties
Extraction Agent
I built this tool to turn invoices from PDFs and images into clean, structured data. It reads the document, extracts key invoice details, validates the result, and lets the user ask questions about the invoice.
- Works with PDFs and images
- Handles text and scanned documents
- Validates extracted invoice data
- 01PDF / image
- 02Input routing
- 03Extraction
- 04Pydantic validation
- 05Structured invoice data
- 06Document Q&A
- Invoice fields
- Line items
- Validated JSON
02 · FOUNDER-BUILT PRODUCT
Financial guidance built around the immigrant experience.
I founded MARVODYN to make the U.S. financial system easier to understand and navigate for immigrants. The live product brings together beginner education, financial-product research, and comparison content across banking, credit, taxes, saving, investing, loans, and international money.
Founder & Product Engineer- Beginner financial guides
- Financial-product comparisons
- Qualified guidance for changing requirements
Voice Agent
A real-time AI voice agent designed to listen, reason, use tools, and respond naturally.
03 · PRODUCTION ENGINEERING → APPLIED AI
From production databases to AI systems
I started in production database engineering, where reliability, incident response, automation, performance, and safe change management were part of the job. That experience now shapes how I build AI systems: with clear boundaries, observable behavior, validation, and production reliability in mind.
Production DBA → Backend & Automation → Applied AI Engineering
04 · EXPERIENCE
The work that shaped how I build.
Before moving into AI engineering, I spent years working in production database environments—handling incidents, performance problems, automation, deployments, and reliability. That experience now shapes how I build and operate AI systems.
September 2022–Present
Diplomatic Solutions Corporation
Senior Production SQL Server DBAInternal Assignment — Applied AI Engineer, Incident Automation
Built the SQL Server investigation path for an internal AI-assisted incident platform, translating production incident-response procedures into restricted Python diagnostics, FastAPI components, Pydantic evidence contracts, and bounded LLM tool-calling workflows.- Developed restricted diagnostic operations for SQL Server blocking, active requests, connection pressure, and database health.
- Designed application-controlled evidence records with verified identifiers and provenance so generated reports reference collected evidence instead of unsupported model claims.
- Integrated database diagnostics into a bounded AI workflow where the model selects approved tools while application code controls execution, permissions, validation, usage limits, and failure handling.
- Added safeguards including strict tool schemas, bounded tool/model usage, human approval, current-state revalidation, and post-action recovery verification.
Production Database Engineering
Support 24x7 mission-critical SQL Server environments across on-premises and Azure infrastructure, including 50+ SQL Server instances, 200+ databases, and databases up to 4 TB.- Serve as a senior database escalation resource during P1/P2 incidents involving database failures, application connectivity, performance degradation, resource pressure, and service availability.
- Perform root-cause analysis for blocking chains, deadlocks, long-running queries, CPU and I/O pressure, TempDB contention, connection exhaustion, wait statistics, and execution-plan regressions.
- Manage Always On Availability Groups, backup and recovery, failovers, HA/DR, monitoring, and recovery operations.
- Automate operational workflows using PowerShell and T-SQL with GitHub, Azure DevOps, CI/CD, and Octopus Deploy.
July 2021–August 2022
Emitek
ETL Engineer / SQL Server DBABuilt and supported Python and SSIS ETL pipelines integrating SQL Server, structured files, XML, APIs, and other business data sources.- Python and Pandas data-processing workflows.
- Query tuning, indexing, execution-plan analysis, and database optimization.
- Python, SQL Server Agent, and T-SQL scheduling/monitoring, including troubleshooting pipeline failures, API issues, and data inconsistencies.
05 · WHAT I BUILD
What I build and how I think about it.
I build AI applications, backend services, agent workflows, and reliable production systems. My focus is not just getting a model to respond—it is building the surrounding system so the result can be tested, validated, monitored, and safely operated.
AI Applications & Backends
FastAPI services, structured outputs, request validation, provider integrations, typed schemas, and production-ready API design.
Agent Workflows
Planning, tool use, orchestration, evidence handling, human approval, and controlled multi-step AI workflows.
Reliability & Evaluation
Testing, logging, observability, failure handling, CI/CD, evaluation, rollback thinking, and production-safe design.
Databases & Production Systems
SQL Server, PostgreSQL, performance troubleshooting, incident response, automation, HA/DR, and production reliability.
06 · CREDENTIALS
Credentials that support the work.
Focused certifications that reinforce my AI, cloud database, and security foundations.
EARNED CERTIFICATIONS
Earned certifications
Claude Certified Associate – Foundations
Anthropic
Demonstrates practical understanding of Claude workflows, prompting, configuration, tool use, and responsible AI system usage.
Verify credential ↗Microsoft Certified: Azure Database Administrator Associate (DP-300)
Microsoft
Validates administration of SQL Server and Azure SQL solutions across security, performance, availability, and migration.View certification ↗CompTIA Security+
CompTIA
Validates foundational cybersecurity knowledge across threats, architecture, operations, and risk.View certification ↗CURRENTLY STUDYING
Claude foundations
- Claude Certified Developer – Foundations
- Claude Certified Architect – Foundations
07 · EDUCATION
Formal education.
2026
Bachelor of Science in Information Technology
Western Governors University
2021
Associate Degree
Montgomery College08 · ARTICLES
What I'm learning, testing, and thinking about.
I write about the questions I run into while learning and building AI systems—from security and agent design to the way I study new tools and technologies.
AI AGENTS · SECURITY
We're Giving AI Agents Tools, Memory, and Permissions. What Could Go Wrong?
A practical look at the security risks that emerge when AI agents are given tools, memory, and permission to act.CLAUDE · CERTIFICATION
The AI Study Loop I Used to Pass the Claude Certified Associate Exam
How I used Anthropic’s official material, ChatGPT, NotebookLM, and practice questions to understand the concepts instead of just memorizing them.CONNECT
Let's connect.
I'm currently focused on AI and Generative AI engineering roles where production experience, backend systems, and reliable AI application design matter.