Staff Engineer & Consultant Amsterdam
I help teams improve one recurring task at a time: decide where AI helps, build in human review, and give the workflow a clear owner.
10+ years across software, platforms, and distributed systems. I check output quality on real tasks, so changes stick and your team spends less time on rework.
Clear scope Practical handover
00 The signal
Teams need a clear way to use the AI tools they already have on work they do every week.
The team has tools, but people use them differently.
Work comes back for another round of edits.
Useful context gets copied between docs, tickets, and spreadsheets.
A useful demo needs a path into the routine.
01 Real work
Generalized from real engineering and AI-adoption work, without identifying details.
ContextTeams were adopting AI tools independently, with no shared way to judge cost or performance.
What I didI ran controlled experiments comparing models and tooling on real work, then optimized caching to cut redundant spend without touching output quality.
ResultAI token costs dropped by about 30%, with much clearer visibility into where the spend was going.
ContextCode review quality and turnaround depended on who had time to look at a given pull request.
What I didI led the rollout of AI-assisted code review: a bounded pilot with guardrails, then full adoption.
ResultThe system now reviews over 10,000 pull requests a month, and AI code-review costs dropped by about 70% along the way.
02 Services
I bring an engineering approach to AI adoption. The workflow can sit in engineering, operations, or another part of the business. Choose one task, define what good looks like, compare real outputs, and hand over what works.
Use current team tasks to practise giving context, structuring requests, reviewing outputs, and weighing cost against quality.
In the roomUse a real task, then turn the useful parts into examples and a playbook.
Take-home
Worked examples, shared terms, and a short playbook for the next attempt.
A good starting point if you already feel where it hurts but can't fully map it yet. We pick one recurring process, fix it on real work, and hand over what changed.
Good candidatesCase handoffs, recurring reports, or meeting notes turned into tracked actions.
Take-home
A workflow checked on real tasks, concise documentation, and an internal owner.
A good starting point if you want the full picture before changing anything. We look at one recurring task, compare cost, retries, and output quality, and some reviews turn into a sprint on what matters most.
The decisionWhich steps need a better prompt, a simpler process, or no AI at all?
Take-home
A short priority list, changes checked against quality, and a clearer standard for future use.
Each engagement stays focused on one clear piece of work. Length follows the problem, from a single session to several weeks.
03 Worked example
Real example
generalized workflow
Before
People across the organization were building similar AI agents and skills on their own, with no way to know what already existed or worked well elsewhere.
After
A shared marketplace lets anyone search for an existing skill or agent, reuse it directly, and publish improvements back for the next person.
Adapted from a real internal AI marketplace used across hundreds of people, generalized here without identifying details.
04 Approach
We start with a real point of friction and work toward a version your team can explain, use, and see the difference.
Find the recurring task that costs the most time or creates the most rework.
Agree on what a useful result looks like and where judgment stays human.
Test the smallest workable change on your team's real material.
Document what changed and leave someone on the team ready to run it.
I'm an Amsterdam-based Staff Engineer with more than 10 years of experience across software, platforms, and distributed systems. That background means I evaluate AI output the way I'd evaluate any system change: on real work, with clear quality criteria, and with an eye on what breaks.
I focus on choosing changes that are useful, checking whether they work, and making the process manageable for the team.
Independent practice
Amsterdam, NL
06 Book a conversation
Bring a workshop idea or a recurring process you want to improve. A little context is enough to begin.
15 minutes
We'll get acquainted, talk through what your team needs, and see whether working together makes sense.
If it's a fit, I'll send a short proposal with scope and next steps within a few days.