My work, as tools your AI can call.

{  "open": true,  "roles": [    "Founding engineer",    "Product engineer"  ],  "companies": "companies of any size",  "work": [    "taking AI and agentic systems from a promising demo to a mature product that people rely on",    "owning a product from a user's problem through to something people use, and keeping it running after launch",    "making the AI inside a product good"  ],  "location": {    "basedIn": "Australia",    "remote": true,    "willRelocate": true,    "relocation": "anywhere for the right role"  },  "contact": {    "form": "https://stefannio.tech/#contact",    "linkedin": "https://linkedin.com/in/stefan-nio"  },  "detail": [    "He is open to roles at companies of any size, as long as the work is taking AI and agentic systems from a promising demo to a mature product that people rely on. Founding engineer and product engineer are the titles closest to how he works, and he would rather own an outcome than a single layer of the stack.",    "Two kinds of work draw him, and at bizsec.io he found them equally rewarding. The first is owning a product from a user's problem through to something people use, and keeping it running after launch. The second is making the AI inside a product good: choosing current models, agents, and tooling where they improve the output or a decision, and leaving them out where plain code is more reliable. He wants to keep maturing how agents work inside real products. That means giving an agent clear, small tasks, a codebase it can read, and checks it can run to prove its own work, then handing it more as it earns that trust. He already works this way at bizsec.io, where agents write and review code inside limits he set.",    "He works remotely from Australia and will relocate anywhere for the right role."  ]}

{  "query": "MCP",  "terms": [    "mcp"  ],  "matches": [    {      "source": "case-study",      "slug": "this-site",      "section": "reflection",      "heading": "Reflection: My worst estimate on this project was off by a factor of six.",      "url": "https://stefannio.tech/work/this-site#reflection",      "next": {        "tool": "get_case_study",        "arguments": {          "slug": "this-site"        }      },      "excerpt": "…the guess, but I now measure before I plan around a number. Next is an MCP server at `/api/mcp`, so an agent can ask for my projects and skills as structured data instead of scraping this page. It will live in a…",      "score": 2    },    {      "source": "case-study",      "slug": "bizsec",      "section": "stack",      "heading": "Stack: Boring where it can be, strict where it matters.",      "url": "https://stefannio.tech/work/bizsec#stack",      "next": {        "tool": "get_case_study",        "arguments": {          "slug": "bizsec"        }      },      "excerpt": "…Anthropic models through the AI Gateway, Stripe, Upstash for rate limits, PostHog, Attio for CRM sync, and Vitest. bizsec.io also ships an MCP server, with its scope set by ADR 0011, so agents can reach the product too.",      "score": 1    },    {      "source": "skills",      "section": "agents-and-llm-systems",      "heading": "Agents and LLM systems",      "url": "https://stefannio.tech/about#skills",      "next": {        "tool": "get_skills",        "arguments": {          "area": "agents-and-llm-systems"        }      },      "excerpt": "Anthropic API, Claude, MCP, Vercel AI SDK, AWS Bedrock, AgentCore, Strands, LangGraph, LangChain, Prompt caching, RAG and pgvector",      "score": 1    }  ]}