"The Best Developer in The World" - based on my wife's ranking

I'm an AI Architect. I design and ship AI systems that hold up in production — not demos. My work sits where the hard problems live: agentic systems at scale, agent and LLM evaluation, AI compliance and governance, and fine-tuning models for real workflows with real accountability.

I think about AI the way an architect thinks about a building: what happens under load, what fails safely, what can be measured, and who owns it a year from now. That means multi-agent orchestration you can actually reason about, evaluation harnesses that catch regressions before users do, guardrails and policy baked into the system rather than bolted on, and fine-tuned models measured against clear baselines.

I build on the Model Context Protocol (MCP) to connect agents to tools, data, and systems securely — and I've spent years proving that JavaScript/TypeScript is a first-class language for serious AI engineering, not just Python. Two decades of large-scale frontend and platform architecture is what lets me ship these systems end to end.

What I architect:

• Agentic Systems at Scale — multi-agent orchestration, planning, tool use, memory, and failure recovery
• Agent & LLM Evaluation — eval harnesses, LLM-as-judge, regression suites, offline + online scoring
• AI Compliance & Governance — guardrails, policy, auditability, safety, and responsible-AI controls
• LLM Fine-Tuning — dataset curation, training, and rigorous before/after evaluation
• RAG & Knowledge Systems — retrieval pipelines, GraphRAG, and grounding at scale
• MCP & AI Tooling — secure, observable connections between agents and real systems

I believe:

• If you can't evaluate it, you can't ship it — evals are the architecture, not an afterthought
• Compliance and safety are design constraints, not paperwork
• Agentic systems live or die on observability and graceful failure
• JavaScript/TypeScript is a first-class language for AI, not just Python
• Good architecture is what makes AI reliable, scalable, and accountable

I speak at conferences worldwide about AI architecture, agents, evaluation, and safety:

Upcoming:

• 🇩🇪 WeAreDevelopers World Congress 2026 — Berlin, Germany
• 🇩🇪 DWX 2026 — Nuremberg, Germany

Recent Conferences:

• 🇮🇱 Reversim Summit 2025 — Tel Aviv, Israel (Oct 2025)
• 🇨🇿 DevConf.CZ 2025 — Brno, Czechia (Jun 2025)
• 🇵🇱 DevoxxPL 2025 — Kraków, Poland (Jun 2025)
• 🇬🇷 CityJS Athens 2024 — Athens, Greece (Nov 2024)
• 🇵🇹 NDC Porto 2024 — Porto, Portugal (Oct 2024)
• 🇸🇬 CityJS Singapore 2024 — Singapore (Jul 2024)
• 🇭🇷 Web Summer Camp 2024 — Opatija, Croatia (Jul 2024)
• 🇺🇸 Visual Studio Live! @ Microsoft HQ 2023 — Redmond, WA (Jul 2023)

→ Full history at danduh.me/conferences

• 🔥 Your Agent Failed. You Blamed the Model. You Were Wrong. — agent reliability is an evaluation and harness problem, not a model problem
• Big Model vs Big Harness: The Debate That's Actually Shaping AI Products — where real leverage in AI products comes from
• Knowledge Is the Infrastructure. Everything Else Is Just Tooling. — RAG, GraphRAG, and grounding as the core of AI systems
• Gen-AI on Localhost: Prompt, MCP, Fine-Tune and RAG&Roll — end-to-end agentic AI, fine-tuning, and RAG on your own machine (workshop)
• Prompts Are Code. Start Treating Them That Way. — prompt versioning & evaluation in production
• MCP Security: How Your Friendly MCP Tool Might Betray You — agent tooling & the attack surface
• AI in Your Browser: Chrome's Built-In LLM — on-device and hybrid inference
• From Idea to Production with AI: Build a Full-Stack App — with VS Code, Copilot, or any IDE (workshop)
• Generative UI: The Future of Frontend — LLMs as the runtime for adaptive, intent-driven UX

→ Full speaker profile & booking: sessionize.com/danduh

I write about agentic systems, the Model Context Protocol, prompt engineering, evaluation, and AI engineering in JavaScript/TypeScript.

🔥 Recent Articles:

• AI Is Building Code Nobody Can Maintain. You're Next. — Jan 2026 · GraphRAG & "AI amnesia"
• You Can't Make Your Company AI-Native Without Dealing With This First — Apr 2026
• Prompt Versioning: The Survival Tool Every Prompt Engineer Needs — Aug 2025
• MCP Servers: Powerful Allies or Sneaky Threats? — Jul 2025
• Stop Worshipping Python: Why JavaScript is the Real MVP for AI Integration — Jan 2025
• Structuring AI Response for Better API Alignment — Jan 2025
• Your Browser Just Grew a Brain: Local-First AI with Chrome's Built-in APIs — Sep 2025
• Generative UI: Smart, Intent-Based, and AI-Driven — Feb 2025
• 10 Advanced TypeScript Features Every Developer Should Know in 2025 — Aug 2025

→ More at danduh.me/articles and Medium

Current Focus:

• Agentic Systems at Scale — orchestration, planning, and memory for multi-agent workflows that stay reliable under load
• Agent & LLM Evaluation — building eval harnesses, LLM-as-judge scoring, and regression suites that gate releases
• AI Compliance & Governance — guardrails, audit trails, and responsible-AI controls as first-class architecture
• LLM Fine-Tuning — dataset curation, training, and disciplined before/after evaluation
• GraphRAG & Knowledge Systems — retrieval and grounding that survive scale and change
• MCP Tooling — secure, observable ways for agents to reach real systems

Recently Published:

• AI Is Building Code Nobody Can Maintain. You're Next. — on GraphRAG and "AI amnesia": what happens when AI builds systems nobody remembers how to own

Technical Depth:

• Evaluation-first delivery — treating evals and observability as the backbone of every AI system
• AI in TypeScript/Node — proving JS is a serious platform for agents, RAG, and tooling
• On-Device & Hybrid AI — Chrome Built-In AI APIs and local inference

Impact:

• Speaking — practical lessons on agents, evaluation, and AI safety at conferences worldwide
• Writing — articles on agentic AI, MCP, and evaluation on Medium and ITNEXT
• Community — helping engineers build AI that's measurable, safe, and production-ready