Autonomous AI agent systems — Harness & Agentic Engineering
Solytics GmbH · Remote — AI / SaaS
Harness engineering · Agent engineering · AI system architecture · Full-stack engineering · DevOps
Building and running autonomous LLM agent systems as a harness engineer — the engineering around the model, not the model itself. The harness is the product: the agent loop and its stop conditions, tool use over MCP, layered memory (durable facts, running context, session state), approval gates for anything irreversible, quality gates that must pass before an agent may ship, observability over every run, budget and rate-limit control with model routing and fallback, plus recovery when a run is killed mid-flight. A local multi-model proxy keeps the whole thing independent of any single provider. Two systems in production: a self-directing AI business partner that plans, writes code, tests and deploys around the clock, and Fabrik (fabrikhq.com), a multi-agent platform for autonomous software delivery with parallel worker agents. Event-driven orchestration over Telegram, GitHub and systemd; several AI-built SaaS products shipped end to end.