Rely on a Rust and AI expert to bring intelligent agents to production, safely and under control
$AI Software: safe, governed agents
What do I offer?
I build AI agents and LLM-based software ready for production: language model integration, developer tooling and intelligent automation. I build the orchestration layer, the tools and the infrastructure around the AI, mainly in Rust.
Why governed agents?
An AI agent does not just make mistakes: it can exfiltrate data, spend money or modify production systems. I design governed agents: every protected action is checked against a policy, routed through a local enforcement point and recorded in an audit trail. An agent can only do what it was allowed to, and every decision stays visible.
In which domains?
Copilots and assistants for DevOps and operations, infrastructure automation, agents that propose commands with human approval (human-in-the-loop), investigation and remediation pipelines, AI backends and LLM integrations inside your products.
Why work with me
I combine years of experience in Rust and distributed systems with the development of reliable AI software. I designed a governed runtime for agents (OpenFirma) and a terminal with an integrated AI agent for DevOps (Infraware Terminal): I know what it takes to bring AI to production with safety, performance and control.
My AI projects
smista.ai
Author of smista.ai, a local-first agent and CLI that routes every stage of an AI workflow to the most suitable model through deterministic, configurable policies: routing never depends on the judgment of an LLM and every decision is explainable through a trace. Built entirely in Rust.
Go to the site
OpenFirma
A governed runtime for AI agents: it starts the agent with a profile, routes outbound traffic through a local enforcement process, checks every protected action against the policies and records the outcome. The agent can only do what it was allowed to.
Go to the siteInfraware Terminal
A terminal emulator with an integrated LLM agent for DevOps assistance. Prefix a command with `?` to ask in natural language: the agent proposes shell commands for your approval before running them, with incident investigation and Docker sandboxing.