Commercial Analytics & Applied AI Leader
I work where commercial numbers lose credibility and AI buying gets fuzzy: in forecast calls with weak stage discipline, in pricing files without control logic, in margin narratives distorted by FX, rebates, or attribution noise, in customer service teams trying to automate poor upstream processes, and in software evaluations where nobody has challenged what the vendor is actually selling. Based near Stuttgart, I rebuild the decision logic, review cadence, and operating model that let leaders act with less noise and more control.
The differentiator is operator range. I have been writing machine learning in TensorFlow for 10+ years while also owning sales, forecast, margin, service, and execution problems in the real business. That combination is what makes me useful when a team has to decide between buying a vendor tool, building with the OpenAI API, fine-tuning a narrower system, or keeping critical logic in-house.
For recruiters and hiring managers
Best-fit role families: Head of Analytics, Director of Analytics, Head of Commercial Analytics, and BI & AI Transformation Lead. For leadership enquiries, review the leadership resume or use the contact path below — consulting buyer engagements (diagnostic reviews) remain the primary service offering.
Selected Delivery Stack
The technology comes after the commercial problem is clear. This stack exists to make messy financial and commercial data more reliable, more auditable, and more usable inside live decision workflows with owners, review logic, and accountability. It is build depth, not course-level familiarity.
Forecasts, pricing, margin, CRM, and executive reporting each need different controls. The stack reflects that.
I favor tools that support traceability, validation, and safe rollout rather than black-box novelty.
Every choice is shaped by what leaders actually need to trust and act on each week.
Power BI carries the decision surface, Python handles repair and orchestration, TensorFlow handles hands-on ML implementation, and LLM APIs are applied only where narrative, retrieval, or controlled automation adds measurable value.
Controlled narrative, commentary, and document workflows
Executive reporting, semantic models, and margin walks
Automation, validation, APIs, and analytics engineering
Experiment tracking across 19+ training runs
10+ years building with TensorFlow since 2016
ML model serving and TelegramNotifications microservice
Semantic search for memoir pipeline and doc retrieval
Live automated stock trading execution
Machine Learning · ML Ops
End-to-end automated trading platform built from hands-on TensorFlow ML code, IBKR live execution, Databento L2 order book data, and W&B experiment tracking across 19 training campaigns.
Challenger Flow
The goal is not to ask buyers what they want to buy. The goal is to help them see the commercial problem more clearly than the vendor or internal project sponsor has framed it so far.
Power BI, Excel, CRM, Copilot pilots, vendor demos, and pressure from leadership to “do something with AI” are already in the room.
The real issue is weak decision logic, poor ownership, broken handoffs, or software choices made before the workflow is understood.
Manual reporting hours, duplicated review work, avoidable support contacts, poor vendor fit, and unmeasured API costs can consume far more value than the licence line suggests.
Meetings start by debating numbers. Support teams inherit worse escalations. Leaders approve AI pilots that look impressive but never become operating capability.
Start with one decision, one workflow, one owner, one cost model, and one control path. Then choose the right software or build pattern around that.
I help diagnose the real decision problem, evaluate the tooling honestly, and then scope the right diagnostic, diligence sprint, workflow build, or enablement path.
Teach First
The strongest first conversation is usually not about the tool. It is about the hidden workflow, governance, and commercial logic that decide whether the tool will actually create value.
That distinction changes data exposure, cost at scale, maintainability, differentiation, and whether the product can survive real production conditions.
Good diligence starts with architecture, permissions, failure modes, and operating cost, not the demo alone.
Many teams try to automate the visible ticket while the real cost sits upstream in poor documentation, unclear invoices, missing order visibility, or weak escalation design.
The safer first step is often an internal copilot, avoidable contact analysis, and cleaner handoffs before customer-facing AI.
Some workflows are best served by off-the-shelf software. Others need a controlled OpenAI API workflow, a thin orchestration layer, or tighter in-house control because the risk and integration burden sit in your own environment.
I help teams choose the right level of customization before they commit to the wrong operating model.
Tailored By Role
Challenger selling is not generic provocation. The reframe has to land in the language of the person who owns the consequence.
CEO / MD
The bigger risk is letting competitors redesign decision workflows faster while your business keeps funding noise, duplication, and AI theatre.
CFO / Finance
The hidden cost usually sits in review work, bad outputs, untracked usage, and weak benefits capture rather than the first invoice.
Sales / Commercial
The aim is better next actions, stronger forecast discipline, pricing control, and cleaner account prioritization, not more reporting surface area.
Operations / Service
Before you launch a chatbot or service agent, check avoidable contacts, knowledge quality, escalation design, and exception rate.
BI / IT / Data
The commercial value usually rises or falls on metric definition, retrieval quality, permissions, maintainability, and workflow adoption after the build.
Lead Magnets
These are practical commercial tools, not generic ebooks. Use them when leadership needs faster clarity on vendor diligence, build-vs-buy logic, and the right next decision.
Vendor diligence
Use this before you approve a pilot or sign a contract. It forces the architecture, data-flow, permissions, failure-mode, and production-cost questions most demos skip.
A polished demo does not tell you whether the workflow can absorb the tool, or whether the review burden simply lands somewhere else.
Need the full teaching context first? Read the guide in the blog.
Build vs buy
Use this when the real question is where the workflow should live: in vendor software, a lean API workflow, a retrieval-oriented system, or a higher-control internal build.
The wrong control model creates switching cost, governance drag, and implementation waste long before the team notices the mistake.
Want the reasoning behind the matrix? The companion article explains how to decide before procurement starts.
Jan 2026 – Present
Medline, Stuttgart
Oct 2022 – Jan 2026
Medline, Germany (Hybrid)
2012 – 2022
Medline, Australia & UK
What's Next?
If forecasting, pricing, margin, CRM quality, or dashboard sprawl are slowing decision-making, or if you are evaluating AI software, support automation, or a first agent workflow, start with a short fit conversation. The usual next step is a fixed-price Diagnostic Review once the problem is clear enough to warrant it.