Tax and Legal Process & Automation Specialist

• Map existing Tax & Legal processes with practitioners and break them down into logical steps, inputs, decision points and outputs
• Turn practitioner expertise into structured logic: decision rules, data models and extraction specifications the engineering team can build on
• Write, test and iterate on prompts for large language models, and compare approaches using data rather than gut feel
• Design and run systematic evaluations: build test sets, measure accuracy and consistency, find failure patterns and propose fixes
• Work closely with developers on solution design, testing and releases to deliver end-to-end solutions
• Support practitioners in piloting and adopting new solutions; gather feedback and turn it into improvements
• Document workflows, test results and design decisions so solutions can be reviewed, governed and reused across countries

• A Bachelor's or Master's degree in engineering, natural sciences, mathematics, computer science, econometrics or a similar analytical field
• 0–5 years of experience; recent graduates with relevant projects, research or internships are welcome
• Strong analytical and structured thinking: you enjoy taking something messy and making it logical
• Genuine interest in AI and how it changes knowledge work; hands-on experience with LLMs, prompting or AI tools is a plus
• Basic programming or data skills (for example Python, SQL or advanced Excel) are a plus, not a requirement
• Eagerness to learn a new domain and comfort asking questions of senior experts
• Clear communication in English, written and spoken
• Comfort working in an international team where priorities evolve as solutions mature

The Global Development AI team sits within Global Tax & Legal at Deloitte Netherlands. We are an international team that builds AI solutions and takes them from first prototype to tools practitioners use on live engagements.

We are building an AI workflow platform for professional services delivery. It combines AI-based extraction of facts from documents with a transparent rule engine, so every output can be traced back to its source.