ABOUT US

Agentic coding tools such as Claude Code, GitHub Copilot, Cursor and Codex are changing how our engineers work, and new capabilities arrive faster than we can currently absorb them. 

You will join a small existing team that already owns this area, with established tooling, working relationships across engineering, and colleagues who know the landscape. You will not be starting from scratch or figuring it out alone. What we need is the capacity and depth to keep pace: teaching teams what has become possible, building it into our templates and golden paths, supporting teams directly when things do not work, and assessing which developments are worth adopting. 

This work sits directly on the Accelerate lever of our “Winning Ways” strategy — enabling people to use AI effectively in their own area of work, and prioritizing the applications whose value has been proven in fast, focused pilots. For software engineering, that is this team's remit.

RESPONSIBILITIES

This position suits an engineer with an active interest in how agentic coding changes software work, rather than a general interest in AI. 

When a new model or agent feature ships, you try it within days, against your own hard problems rather than a toy example — a legacy refactor whose shape you already know, a test suite that resists, a migration you have previously done by hand. Your views come from that, not from reading about it, and you can distinguish a real capability improvement from a well-produced demo. The second-order question interests you as much as the result: if this holds, what does it change about how we review code, structure repositories, write specifications and run CI? 

You will probably recognize yourself in the following: 

• You write and iterate on your own agent configuration — instruction files, custom commands, subagents, MCP servers — rather than accepting the defaults. 
• You know how differently these tools behave on a large, old, inconsistent codebase than on a greenfield project, and you have views on what makes a repository navigable to an agent. 
• You have reviewed a lot of agent-written code and can say where it reliably goes wrong: changes that pass the tests but are wrong, plausible refactors that quietly lose behavior, work that looks finished and is not. 
• You care about the state of the codebase, not only about throughput. You have seen how quickly quality can erode when generation is cheap and review is not. 
• You enjoy making other engineers more effective. Teaching is a substantial part of this role, not an occasional obligation. 
• You already follow this field closely in your own time. What this role adds is the mandate, the budget and the working hours to do it properly. 

Responsibilities 

Technical enablement and forward-deployed support (40%) 

• Teach engineers how to use new capabilities effectively — not only that a feature exists, but what it changes in day-to-day work, demonstrated on our own code. 
• Work directly with teams on their real problems: pairing, debugging agent workflows, and investigating why a tool that performs well in a demo does not perform on our codebases. 
• Run sessions, office hours, demos and written guides, and turn recurring problems from the field into permanent improvements. 

Developer platform and templates (25%) 

• Equip our templates and golden paths with agent harnesses: agent instruction files, MCP server configuration, custom commands and subagents, tool permissions and sensible defaults, so services work well with agentic tools from day one. 
• Define and track the DevEx metrics that show whether these changes have real impact — adoption, cycle time, delivery throughput, developer-reported friction — so template changes and model swaps can be assessed on evidence. 
• Make newly approved models available to teams through our existing tooling layer. 

Model and feature evaluation (20%) 

• Test new models as they are released, open-source and commercial, and establish how their capabilities compare to the frontier models we already use. 
• Evaluate against tasks that reflect our actual work rather than public leaderboards, and be specific about where a cheaper or open model is sufficient and where it is not. 

Market and concept validation (15%) 

• Follow the landscape and distinguish substantive developments from noise, covering tools, protocols, agent patterns and workflows. 
• Run structured, time-boxed evaluations of promising concepts in our own context, and document what was tested, what you recommend, and what we should not pursue for now. 

QUALIFICATIONS

• Practicing engineer. You write production-quality code and have been trusted with real systems. We are not prescriptive about the stack: deep knowledge of any modern open-source language is what counts. Python and TypeScript come up most often here, but strong Go, Rust, Kotlin or comparable backgrounds are equally welcome. 
• Daily use of agentic coding tools — Claude Code, GitHub Copilot, Cursor, Codex — with a clear understanding of their limitations as well as their strengths. 
• Familiarity with the underlying mechanics: LLM APIs and tool calling, MCP, context management, agent and prompt design, token economics, common failure modes. 
• Engineering fundamentals: Git, CI/CD, containers, at least one cloud platform. 
• An evidence-based approach. You can design a test that answers a question, and revise your view when the results say you should. 
• Accountability. You give and take ownership rather than working under close supervision: you identify where the improvement potential lies, develop your own concept for it, and carry it through to adoption. Much of this work is not specified in advance — you will define it more often than you receive it, and you are measured on the result rather than the effort. 
• Speed. You would rather put a working improvement in engineers' hands this month and refine it than design the complete solution first. You time-box your evaluations, and you focus on what matters most instead of doing a little of everything. 
• Improvement mindset. You argue for the best solution rather than the easiest compromise, you give candid and constructive feedback rather than empty phrases, and you actively bring in outside perspective — from the market, from open source, from other engineering organizations — rather than working only from how we do things today. 
• Strong teaching and communication skills. Making other engineers effective is a core part of the role, not a side effect of it. You build momentum for new ways of working rather than skepticism. 
• Fluent English, written and spoken. 

Nice to have:

• Public work we can review: a GitHub profile, open-source contributions, an MCP server, a tool, articles or talks. 
• Experience with internal developer platforms such as Backstage, or with scaffolding and template systems. 
• Experience with DevEx and developer productivity metrics (DORA, SPACE, DX Core 4 or similar). 
• A background in which other engineers were your users: forward-deployed or solutions engineering, developer relations, consulting, platform teams, or teaching. 

BENEFITS

• A secure work environment because your health, safety and wellbeing is always our top priority.
• Flexible work schedule and Home-office options, so that you can balance your working life and private life.
• Learning and development opportunities
• 25 holiday days per year
• 5 additional days (readjustment)
• A collaborative, trustful and innovative work environment
• Being part of an international team and work in global projects
• Relocation assistance to Madrid provided

At BASF, the chemistry is right.

Because we are counting on innovative solutions, on sustainable actions, and on connected thinking. And on you. Become a part of our formula for success and develop the future with us - in a global team that embraces diversity and equal opportunities irrespective of gender, age, origin, sexual orientation, disability or belief.

Contact

Do you have any questions about the application process or the position? Please reach out to Felipe Bianco (felipe.bianco@partners.basf.com)