Harness Engineer

Anitech Solutions is hiring a Harness Engineer to build and improve TypeScript-based platforms that use AI agents under human control. The role also configures and runs the harnesses on client engagements, ensuring quality, security, governance, and measurable results.

Responsibilities:

• Extend the deterministic orchestrator that owns control: agents run bounded tasks, and only the orchestrator decides what happens next
• Build and tune the agent loop: the context each agent sees, the tools it may hold, and the model class it runs on
• Connect the harnesses to the systems clients use: issue trackers, Git hosts and CI on GitHub and GitLab
• Build the React operator screens used to start, watch, approve and replay runs
• Set up the harness on each new client estate: repository discovery, coding standards, gates and configuration
• Run the harness on the client's backlog, working alongside their engineers, product owners and approvers
• Own the quality of what ships: review merge requests, explain the audit trail, and fix what the gates catch
• Show clients how the governance works, so their own people can approve with confidence
• Build the evaluation and replay tooling: scenario suites, regression runs and replay of real runs from the event log
• Add metrics and cost metering so every ticket reports cycle time, tokens and cost
• Own the sandbox: pinned images, no network, read-only file systems and resource limits
• Redact secrets and client personal data before anything is written to disk or logs
• Hold the line that an agent never merges its own work and never records a human approval
• Turn every slowdown, workaround and failure on a client engagement into a harness fix, with a test that proves it
• Cut the set-up time for each new client by making what you configured once reusable
• Write ADRs for decisions that constrain future work and review other people's
• Keep one gate that runs the same locally and in CI and never weaken it to get a change through
• Adhere to company and customer policies, processes and standards, including our ISO/IEC 27001 Information Security Management System and data protection requirements, and promptly report any security concerns or incidents