David O.

I build AI agents that hold up in production. I'm a software engineer with 12+ years of experience, with a current core focus and expertise on Multi-Agentic AI Systems. I've designed and delivered AI solutions for the United Nations and for startups at many stages. I'm Top Rated on Upwork with 100% Job Success, and nearly all of my recent contracts have been AI and agentic roles. If you need agents that route correctly, use tools safely, remember context, stay within budget, and can be traced and tested, that is exactly what I do. WHAT I BUILD • Multi-agent systems: triage-to-specialist handoffs, manager/orchestrator patterns, parallel agents, deterministic chains, and LLM-as-router • Tool-using agents: function tools, hosted tools (web search, file search, code interpreter, computer use), agents-as-tools, and solid tool error handling • MCP servers and integrations: custom MCP server development, local and remote servers, tool filtering, and approval flows • Memory and context: sessions, conversation state, dependency injection through run context, and long-term memory • Voice and realtime agents: speech-to-speech and voice pipelines • Human-in-the-loop workflows: approval gates, interruptions, and pausing and resuming runs HOW I MAKE THEM PRODUCTION-READY • Guardrails: input, output, and tool guardrails, tripwires, blocking vs. parallel checks, and prompt-injection defense • Observability: built-in tracing, custom spans, and Langfuse and Logfire integration • Evaluation: trace grading, eval datasets, and regression tests, so a prompt change never silently breaks behavior • Reliability and cost: retries, error handling, max-turn limits, and latency and token-cost control • Structured outputs: typed, validated responses with Pydantic • Streaming UIs: event streaming, run hooks, cancellation, and frontend integration • Deployment: FastAPI services, containerized and cloud-ready FRAMEWORKS AND TOOLS OpenAI Agents SDK • Google ADK • A2A Protocol • AWS Strands Agents • Agno • CrewAI • Mastra • MCP • Pydantic • FastAPI • Langfuse • Logfire • Python • TypeScript WHY CLIENTS WORK WITH ME • Engineer first: I have 12+ years across the full SDLC, including architecture, clean code, testing, CI/CD, and maintenance. Good AI sits on top of solid engineering, not in place of it. • Framework-agnostic: I choose the stack that fits your problem, not the one I happen to prefer. • Clear communication: you'll always know what's being built, why, and what it will cost to run. • Proven results: UN-scale projects, startup launches, Top Rated status, and 100% JSS. BEST FIT • Building an AI agent or multi-agent product from scratch • Rescuing a prototype that fails under real-world use • Adding MCP servers or tool integrations to your platform • Making existing agents observable, testable, safer, and cheaper to run Send me a message about what you're building, and I'll reply with how I'd approach it.