Available · Q4 · 2026

Production-ready agentic systems for web & mobile.

Available for projects

I'm Tim Witter, a full-stack engineer in Vienna working on AI systems. I design, build, and ship agents, LLM pipelines, and RAG — including the infrastructure that runs them. Rust and TypeScript at the core, .NET where it earns its keep. I work end-to-end: from the data and inference layer to the interfaces your users actually use. Enterprise experience, hands-on with current AI tooling. Built to survive production, not just the demo.

Start a project See what I build

Based · Vienna, AT Pillars · Rust · TypeScript Also · .NET · NixOS · Web3 Response · within 48h

Agents in production LLM pipelines RAG over messy data Rust services Next.js apps Realtime mobile Vector + graph retrieval Eval harnesses Inference infra NixOS reproducibility Web3 settlement Workflow automation Agents in production LLM pipelines RAG over messy data Rust services Next.js apps Realtime mobile Vector + graph retrieval Eval harnesses Inference infra NixOS reproducibility Web3 settlement Workflow automation

What I build.

Four overlapping practices. Most projects touch all four; they rarely separate cleanly.

01 · capability

Agentic systems

Multi-step agents that plan, call tools, and recover from failure. Built with eval-first loops so they keep working when models, data, or requirements move.

LangGraph MCP Tool use Guardrails

02 · capability

LLM & RAG pipelines

Retrieval over messy, real-world corpora. Hybrid search, structured grounding, and prompt + index versioning that holds up in production.

pgvector Hybrid search Re-ranking

03 · capability

AI infrastructure

The unglamorous half: inference, queues, observability, cost controls, replay. So your model layer behaves like the rest of your stack.

Inference Queues Tracing Cost budgets

04 · capability

Production web & mobile

End-to-end product surfaces. Next.js, Nuxt, Astro, native mobile — wired to APIs you can trust, with the AI parts treated as first-class, not bolted on.

Next.js Nuxt Astro Mobile

The stack — honestly used.

Core languages

Rust TypeScript C# / .NET Python SQL Solidity PowerShell Bash Nix

Frontend

React Next.js Astro Nuxt Vue Angular Tailwind

OpenRouter OpenAI Anthropic Google LangGraph MCP A2A pgvector vLLM Local Llama

Backend & data

Node / Bun ASP.NET Postgres Redis Kafka

Infra

NixOS Docker Kubernetes Cloudflare Terraform GitHub Actions

Web3 & misc

Tauri Expo

Agentic Runtime

OpenClaw OpenFang ZeroClaw

Coding harnesses

OpenCode ForgeCode ClaudeCode Codex Copilot CLI

Selected work.

A few representative builds. Full case studies in conversation, not on the homepage.

CreatorPulse

Analytics dashboard for content creators with real-time metrics and audience insights.

React Native · Next.js · Drizzle

MintShot App

Mobile app for capturing and organizing visual inspiration with AI-powered tagging.

React Native · Expo · Neon

FinanceClaw

Personal finance tracker with automated transaction categorization and budgeting tools.

Tauri v2 · Rust · React · Sqlite

How I work.

Four positions I bring to every engagement. Challenge them in conversation; I won't let them erode in the codebase.

Ship the boring half.

Most AI features fail on infra, not on intelligence. Queues, retries, evals, and tracing are part of the deliverable.

Models are dependencies.

Pin them. Test them. Swap them. Production behavior shouldn't change because a vendor shipped on a Thursday.

III

Pick the boring language.

Rust where correctness pays. TypeScript where iteration pays. Everything else is a tool, not a tribe.

Less, but on time.

I'd rather scope a project down than slip it. The version that ships beats the version you imagined.

Let's talk

Have a problem that needs building?

Tell me what you're trying to ship. I read every message and reply within 48 hours. No funnel, no discovery call gymnastics.

hello@timwitter.com