Hire an AI agent developer
Warren IT is the consulting practice of Gabe Warren — a senior engineer who ships production AI agents, MCP servers, AWS Bedrock systems, and custom software in C#, .NET, Blazor, and Python. Bounded engagements that produce a working slice, not a deck.
Atlanta, GA. Remote across the United States. Side projects, fractional, and full engagements.
Tool-using agents that retrieve context, call APIs, update systems of record, draft work for human review, and route exceptions where they belong. Built with Anthropic Claude, the Claude Agent SDK, AWS Bedrock, or your preferred model provider.
Custom Model Context Protocol servers that expose internal APIs, data, and workflows to Claude, Claude Code, and other MCP-aware clients — with proper auth, rate limiting, structured logging, and a thin evaluation harness.
Production Bedrock apps with cost discipline. AgentCore workflows. Nova Act browser automation. Cognito auth, SES email, Lambda, S3, IAM, and CloudWatch wiring that holds up under load.
Blazor portals, ASP.NET Core APIs, MongoDB-backed apps, Python services, and the boring-but-critical operating layer around AI: admin tooling, billing, role-based access, audit trails, and observability.
Stuck deploys, fragile auth, broken integrations, slow releases, and stalled AI proofs of concept. Bounded triage, an honest diagnosis, and a fix path that fits the calendar you actually have.
Ongoing retainer for teams that need senior judgment on AI and cloud architecture without hiring a full-time staff engineer yet. Architecture review, hands-on code, incident help, and hiring support.
Describe the workflow, the current stack, the deadline, and what is blocked. You will get back whether I can help, what I would inspect first, and a realistic next step.
Same-week kickoff is typical for bounded engagements. Larger builds get a written estimate and a calendar slot.
Yes. Mutual NDA is the default for any project review that involves real systems or data.
Yes. Most engagements include working alongside in-house engineers, reviewing code, and pairing on the harder integration points.
Custom foundation model training, vague AI strategy decks, long discovery phases with no build plan, and anything that requires moving money or acting on behalf of unverified third parties.
You do not need a polished spec. A short description, current stack, deadline, and what is blocked is enough for a useful first response.