Forward Deployed Agent Engineer
Working with MING as a Forward Deployed Agent Engineer, you embed inside our client engagements and turn a messy, real business problem into an AI agent that does the job: shipped against real data, validated against the client's own benchmark, and owned end to end. You are the person who sits with a client, works out what is actually broken, builds the agent that fixes it, and proves it works before anyone is asked to trust it. This is where our hybrid-organisation work meets the ground. Agents that hold real roles inside serious industrial, energy, and finance enterprises, built by you, in the room.
The craft here is more specific than shipping fast. It is knowing what should become a reusable skill versus a scheduled job versus left to the model. It is making depth and traceability the agent's default rather than a lucky output, validating against the client's own reference until the agent beats it, and designing how the agent escalates, defers, and holds authority with each stakeholder. The agents you build are measured on accuracy and trust, not demos.
You report straight to Sebastian Mueller, Founding Partner, and build on the platform owned by our Agent Platform Engineer, pushing back the requirements that make it better. Early on you take client context and trust from Sebastian directly. The goal is that you carry engagements he cannot personally sit in.
We offer you
- Real ownership of whole client agents, end to end. You are the person who made them work, and the person clients come to ask.
- Frontier client work, embedding the first agents into demanding enterprises with real outcomes on the line.
- A founder who works the engagements with you, and a platform team that builds the tooling so your time goes to the client's problem, not the plumbing.
- A founding role on a new function, with a clear path to lead the field team.
- A varied client base and a front-row seat to how serious enterprises actually adopt AI.
- Flexible, location-open working.
Your specific responsibilities include, but are not limited to
Decomposition & scoping
You turn a business problem into a buildable agent.
- Translate a client's problem into an agent specification grounded in the outcome, not the tech stack.
- Define responsibilities, decision-authority tiers, data flows, and the measures that prove it works.
- Make the hard scoping calls: what becomes a reusable skill, what is a scheduled job, what stays with the model, and where a human stays in the loop.
Build & ship
You build in the field, fast and well.
- Engineer the skills, prompts, and context that make the client's agent reliable and make depth and sourcing its default behaviour.
- Integrate the agent with the client's real systems, data, and workflows.
- Iterate against real use until the agent is trusted, not just demonstrable.
Validation & evidence
You prove it works before anyone relies on it.
- Validate against the client's own reference and real data. Real-data runs and accuracy gates beat passing tests.
- Catch and purge fabrication, drift, and looks-right-but-isn’t output, in the agent and in your own status reports. "Done" without evidence is not done.
Governance in the field
You design how the agent behaves under authority.
- Set the agent's escalation tiers, autonomy levels, and human gates per stakeholder and per risk class.
- Build the audit trail and disclosure behaviour a client can defend internally.
Client delivery
You are a credible MING technical face in the engagement.
- Hold trust in the room. Judge what to surface and what to escalate.
- Communicate clearly with technical and non-technical stakeholders alike.
- Write clean specs, role briefs, and client-facing artefacts.
Productise back
You make the next engagement faster.
- Turn bespoke client solutions into reusable patterns and runbooks.
- Feed platform requirements and recurring needs back to the platform team.
Our ideal candidate brings
- Evidence-first discipline above all. You verify at the code and output level, you distrust your own "done", and you would never let an unverified claim reach a client. This is the single trait that makes or breaks the role.
- Scoping and validation judgment. You know what to turn into a skill, and you measure an agent against the client's real benchmark, not a vibe.
- Strong skill, prompt, and context engineering. Applied craft, not trial-and-error.
- Genuine business literacy. You start from "what is the outcome, and who owns this agent?", with the stakeholder poise to carry it in front of enterprise clients.
- A working security discipline in the field. Least privilege, credential isolation, never handing an agent more than it should hold.
- High agency in messy environments. You ship where the ground is uneven, with the judgment to run an engagement build largely unsupervised.
Nice to have: B2B enterprise or consulting exposure (industrials, energy, finance); German (a real advantage in our client rooms); experience taking bespoke solutions to repeatable patterns.
Why MING
We are a small, senior team building the operating model for organisations where humans and agents share the org chart. The forward-deployed role is where that stops being a thesis and becomes working software inside a real company, built and proven by you. If you want to ship at the frontier, own outcomes, and define a role that barely exists yet, this is it.
Think this is you?
Send the AR department your work: a link to what you've built, and a few sentences on what you'd build with us. We read everything, and reply to the ones we'd want to work with.
Apply for this role
