Adjust how this site looks and moves. Changes apply immediately to every page.
Settings are saved only in this browser and are never sent to a server. Privacy information
I design and build AI and agent systems, reproducible Linux and NixOS infrastructure, and the network and security layers around them, for developer teams, startups and anyone working on hard technical problems.
Scoped projects. Hourly support. Part of your team.
Based in Vienna · Remote worldwide· On-site in Vienna
How I can help
I work where software meets the machine: agents that run safely, hosts that can be rebuilt from a file, and networks that only allow what they should. That might mean building something new, hardening an existing system, or making it easier to operate.
I build AI systems, LLM integrations and agent workflows with the boundaries they need: isolated execution, scoped tool access, controlled egress and clear ownership of credentials. The aim is agents you can operate, audit and trust.
Agent workflows · Isolated runtimes · Tool & model gateways
Declarative hosts, reproducible environments, and MicroVM or container isolation. I replace hand-built servers with configuration that a team can review, rebuild and roll back.
NixOS hosts & flakes · MicroVMs & containers · Automation
Segmentation, egress control, secrets handling and access policy, backed by a written threat model. I design for the failure you expect rather than the one you hope to avoid.
Threat models · Egress & segmentation · Secrets & access
Command-line tools, services, terminal UIs and APIs in Rust, Go and Python: software that operates systems rather than decorates them. Web interfaces when they serve that system, such as a dashboard.
CLIs & services · Terminal UIs · APIs & integrations
What I’m building
My current open-source work: a NixOS tool that keeps AI agents running around the clock on a headless Linux machine. Each agent lives in its own isolated container, declared in a flake, and Tentaflake is the environment around them all.
DECLARATIVE
Hosts, agents and their runtimes are configuration you can review, rebuild and roll back.
Hermes and ZeroClaw on x86_64 NixOS
ISOLATED
Each agent runs in its own container with its own secrets, skills and persistent workspace. By default it has no direct network access; models and the web are reached through policy.
OCI containers under systemd
ALWAYS ON
Declare an agent once and it runs on a Linux box with no desktop session. systemd restarts it after a crash, and unattended start requires its policy boundaries to be in place.
Rust broker and operator CLI
The agents in the background
An illustration of agents at work in a Tentaflake environment, not a live view.
OPEN SOURCE Pre-1.0 and under active development: interfaces and defaults may still change.
Ways to work together
Bring me a defined project, a specific problem, or a role within your team, whether you are a developer team, a startup or one person building something technically demanding. We agree on responsibilities, priorities, and a way of working.
A DEFINED OUTCOME
I work from an agreed scope and milestones through implementation and handover, with clear communication about progress and technical decisions.
Scope → Milestones → Handover
FOCUSED HELP
I help with targeted implementation, debugging, integrations, and improvements when you need engineering support for a specific piece of work.
An issue, a feature, or an integration
PART OF YOUR TEAM
I join your existing workflow, use your tools and conventions, and take ownership of agreed responsibilities alongside your engineers.
Ongoing work, shared context
About & approach
I’m an independent engineer based in Vienna. I like systems with clear boundaries: where an agent may act, what a host may trust, what a network should allow.
I start by understanding the requirements, the existing code, and the people working with it. I explain trade-offs, make decisions in context, and aim for software that remains understandable after handover.
I’m comfortable owning a defined piece of work and collaborating with the team around it. The tools depend on the problem; maintainability and clear communication stay part of the process.
A good fit: developer teams, startups and individuals building technically demanding systems, such as agent platforms, infrastructure, and network or security tooling. A weaker fit: generic SaaS products that are likely to be commoditised within months. I will say so early.
Rust / Go / Python /TypeScript / Bash
Linux / NixOS/systemd / Tailscale
Have something in mind?
Or copy my address into your email app: hello@timwitter.com
For a project, tell me what you need, your expected timeline, and whether you prefer a scoped project, hourly support, or work with your team. For a technical question, a short description and relevant context are enough.
