AI Harness Engineer

Milestone Systems; Glostrup Kommune; On-site; DKK 900,000 - 1,400,000; Full time

Benefits offered by this job

Development plans

Training resources

Flexible hybrid

Supportive leadership

Cross-functional collaboration

People First culture

Social activities

Job summary

Milestone Systems seeks an AI Harness Engineer (Senior/Principal) to own the AI engineering platform, including gateway, agent loops, context handling, and tool integrations. You’ll define standards and architectures for AI-assisted development and build a robust observability layer across teams.

You will turn existing experiments into scalable internal solutions, establish evaluations and cost-tracking, and guide decisions on build vs buy, while ensuring on-prem and air-gapped deployments.

Qualifications

• Staff engineer or senior IC with distributed production systems experience.
• Hands-on familiarity with AI coding tools (Claude Code, Cursor, Copilot, Codex).
• Experience deploying LLM-based systems in production and managing latency, cost, and reliability.
• Strong fundamentals: diagnosing failures and fixing root causes.
• Budget mindset: manage token and compute spend as a real budget.

Responsibilities

• Own the AI engineering platform: model gateway, agent loops, context handling, tool integrations, caching, guardrails, identity, secrets, and audit trails.
• Define technical standards and reference architectures for AI-assisted development across teams.
• Build the observability layer to trace an agent run end-to-end and diagnose failures.
• Turn internal experiments into scalable solutions for other teams to use.
• Establish evaluations and cost-tracking to assess tool value and performance.
• Inform build-vs-buy decisions, prioritizing product-related components for build.
• Ensure operability in on-prem and air-gapped deployments.
• Collaborate with product teams to extend approach to customer-facing AI features within EU Act and GDPR.
• Mentor engineers and drive adoption across skeptical teams.

Skills

Distributed systems

AI tooling

LLM production

Observability

Security credentials

Cost awareness

Engineering judgment

Tools

Claude Code

Cursor

Copilot

Codex

Job description

AI Harness Engineer (Senior/Principal)

We've had real wins with AI at Milestone, but they've mostly stayed inside the teams that found them. We're looking for someone to take what's working and build it into a platform the whole engineering organization can rely on.

The work sits in the layer around the models: the gateway, agent tooling, context handling, evaluations, guardrails, observability, and cost controls that make AI dependable for real engineering work. You'll own that layer, not just run it, but set the technical standards and reference architectures other teams build against.

The first phase is our internal development environment. From there, you'll bring the same rigor to the AI features going into our products, working alongside the teams that own them.

Part of the job is being straight with us about where AI doesn't help, not only where it does.

On level: We're open at both senior and principal. At senior, we expect you to build and operate this platform. At principal, we additionally expect you to set the technical direction for AI‑assisted engineering across the organization — the person other staff and principal engineers argue with and come away convinced by.

What you'll do

• Own the AI engineering platform: the model gateway, agent loops, context handling, tool integrations, caching and guardrails, along with the identity, secrets handling, and audit trails around them. This is production infrastructure serving the whole engineering organization, with the availability and security expectations that implies.
• Define the technical standards and reference architectures for AI‑assisted development: how teams integrate agents into their workflows, what good looks like for context and evaluation, which patterns we support and which we don't.
• Build the observability layer. Trace an agent run end to end, so that when something fails we can say whether it was the model, the context, the tool call, or the code — and act on the answer.
• Take the experiments already working somewhere in the company and turn them into something other teams can pick up and run with.
• Stand up the evaluations and cost tracking we need to tell whether a given tool is pulling its weight.
• Own the build‑versus‑buy recommendations. We should buy most infrastructure; the parts tied to our products, or to our on‑prem and compliance requirements, are where building may make sense.
• Make it run where we run. On‑prem and air‑gapped deployment is a first‑class design constraint here, not an edge case.
• In the second phase, work with product teams to bring the same approach to customer‑facing AI features, inside EU AI Act and GDPR.
• Get engineers on board the right way — by showing them, pairing with them, and writing things down — so the change outlasts you.

What you'll bring

• Time spent as a staff engineer or senior IC building and operating distributed production systems: API and service boundaries, failure modes under load, observability, and the security work that comes with handling credentials and sensitive source code.
• Real, hands‑on familiarity with today's AI coding tools (Claude Code, Cursor, Copilot, Codex and the like) and the craft around them: context engineering, prompting, caching, tool use, memory, and the rest of it.
• Experience putting an LLM‑based system into production and keeping it there: evaluation, failure modes, latency, cost, and what breaks once real traffic arrives. Internal platforms count as production.
• Strong fundamentals. When an agent fails, you can work out why and fix it rather than just file a ticket.
• A habit of treating token and compute spend as a real budget, and a sense of how to bring cost‑per‑task down.
• Some history measuring this kind of work — adoption, quality, productivity, or ROI — using frameworks like DX Core 4 or DORA.
• The presence to stand in front of experienced, skeptical engineers, have the hard conversation, and still bring them with you.
• A view that AI is a powerful tool but that engineering judgment is what makes the difference, plus enough curiosity to keep up as the field shifts.

Nice to have

• You've shipped agentic features in a product customers actually use.
• Experience building or operating under EU AI Act or GDPR requirements.
• You've run an internal model gateway, LLM monitoring, or evaluation tooling at organizational scale.
• Experience with on‑prem or air‑gapped deployments.

About the team

You’ll join a new team in the Experience Development Organization, set up to put Milestone on the front edge of AI‑assisted development. The team is being built now: you'll work alongside other AI Harness Engineers as it grows, and closely with engineering teams across the organization. [Add current and planned team size, and who the role reports to.]

What we offer

• Individual development plans aligned with your career goals
• Access to training, learning resources, certifications, and professional development
• Flexible hybrid working arrangements
• Supportive leadership and regular feedback
• Opportunities for cross‑functional and international collaboration
• A People First culture built on trust, inclusion, and empowerment
• Social activities and spaces to connect with colleagues across the organization

Practical details

Please note that applying for this position requires a valid Danish work visa.

About us

Milestone Systems is a leading provider of data‑driven video technology software. Our portfolio includes XProtect video management software, Arcules video surveillance as a service, and BriefCam analytics. We amplify what organizations of any size can see, do, and achieve with video.

Guided by our People First mindset — freedom, inclusion, and togetherness — we always consider our employees, partners, and communities in everything we do.

Location

Copenhagen, Denmark