Forward-Deployed Engineer (m/f/d), Junior to Senior

You will not be living out of a suitcase. Travel runs well under 20%. Most of the embedding happens differently: you work remotely but directly inside the client's own systems and infrastructure, their ERP, their data platform and their operational reality rather than a sandbox copy of it. On top of that you run short, high-intensity on-site sessions at the moments that matter, such as kickoff, requirements deep-dives and go-live.

Most of what makes an agent valuable in a pharmaceutical or chemical supply chain happens before any model runs. You sit with planners and operations leads and do real requirements engineering: working out which decision is actually being made, by whom, on what evidence, and where it currently goes wrong. Then you re-engineer the process around it so the data becomes decision-ready, which usually means changing how it is captured, joined and given context rather than simply moving it somewhere new. This is the part most vendors skip, and it is the reason their pilots stall.

• The corporate cortex for an account: the institutional knowledge layer an agent reasons against, so that what the organisation already knows is available at the moment of decision instead of living in one planner's head.
• Skills that encode a client's hard-won planning expertise into repeatable reasoning, so it is applied consistently rather than rediscovered every cycle.
• Agentic teams: specialised agents and subagents decomposed across a dependency graph and coordinated toward a single outcome.
• The human side of the loop. How a planner reviews, approves, corrects and overrides an agent, and what the system does with that correction afterwards. Adoption lives or dies here, and it is a design problem as much as an engineering one.
• Permissioned, structured reach into live systems through MCP, rather than ad-hoc integrations.
• A working map of the client's supply chain system architecture across planning, ERP, execution and data layers, which is usually the first thing nobody has written down.

• Python and TypeScript as primary languages. What we are really hiring for is the judgment to read and reason about any architecture you encounter, whatever language it is written in. We are not hiring for syntax memorisation, that is what Claude Code is for. We are hiring engineering judgment.
• Agentic LLM orchestration with the Anthropic SDK, LangGraph and MCP, deployed against real industrial complexity rather than a demo environment.
• Our own simulation and optimization stack, which is what the agents plan against.
• At our customers: SAP and S/4HANA, Snowflake, Databricks and the wider enterprise data landscape. You meet these on the client side and need to be effective inside them.
• AWS as our primary platform, with Azure or GCP where a client's environment requires it.

• Junior0 to 2 years

  No prior professional experience needed. You work beside senior engineers inside live client systems.

• Mid2 to 5 years

  You have shipped production systems and are comfortable inside a client's existing infrastructure.

• Senior5+ years

  You own client-facing technical outcomes end to end and set the technical direction of engagements.

These apply at every level. How much depth we expect scales with where you join, so read them against the levels above rather than as one fixed bar.

• A degree in Business Mathematics, Industrial Engineering, Computer Science or a related quantitative or technical field, anywhere from Bachelor's to doctorate.
• A requirements engineering instinct that sharpens with experience. You sit with a planner describing a symptom and leave with the actual decision problem written down.
• Comfort with messy enterprise data and system landscapes, whether you are mapping your first ERP export or your tenth. The client environment will not simplify itself.
• Interest in how people work with agents, not only in how agents work. Review, approval and override are where most of the value is won or lost on a project.
• Exceptional problem-solving under real complexity and ambiguity, working inside someone else's systems: supported delivery at entry, owning the engagement later on.
• English is required for the role, and German is a plus for on-site conversations with the client teams whose processes you re-engineer. Travel stays under 20%.

You report into a Principal Consultant or Delivery Lead. From here the path splits depending on where your strengths pull you: deeper into a Senior AI Engineer track, with more technical depth and less client exposure, or toward Principal Consultant, with more client ownership and less hands-on code. Both are real, well-worn paths at adago, not just a line in a job ad.