Last updated August 2026
Four full case studies from work at BMW Group and Siemens Energy, plus two products I shipped on my own. Each one covers the problem, the design decisions, and the measured result, not just the screenshot.
Some artifacts are anonymized due to confidentiality. Where a visual is recreated or a system cannot be shown, the page says so.
BMW GroupA retrieval-augmented assistant built on BMW's internal AI platform, adopted by 12 consultants across international markets and later localized for South Africa and Australia.
Adopted by 12 consultants, around 2-3 hrs/week saved each.
Siemens EnergyA Microsoft Copilot agent grounded in the department's SharePoint, capturing retiring experts' knowledge through structured interviews and measured monthly against real questions.
Answer accuracy improved from ~80% to ~90-95%.
Live productA live interactive map of where data-center and water-infrastructure growth may create electricity demand, built solo as a side project and now used by Siemens Energy colleagues.
100+ public data feeds, 197 countries, daily automated refresh.
Siemens Energy · Open sourceAn open-source market-research agent that reads real customer opinions across Reddit, YouTube, and review sites, rebuilt from the internal review-mining agents I set up at Siemens Energy.
Customer sentiment at scale, categorized automatically.
Real work that is smaller in scope than the four case studies above, kept here rather than stretched into its own page.
Using Power Automate, I built an agent that collects my meeting transcripts, summarizes them, and generates a presentation from the summarized content. A raw recording becomes a shareable deck with no manual work in between. It is a small build, but it shows the pattern I apply everywhere: find a recurring manual workflow, then replace it with a monitored automation using the tools already available in the environment.
In many teams, AI tools exist before people know where they actually help. At BMW I ran a 3-hour workshop for 15 attendees on using the RAG assistant for faster research and grounded first drafts. At Siemens Energy I created a full Copilot usage playbook and delivered a live workshop to around 25 colleagues covering Teams, Outlook, Word, Excel, and PowerPoint. It is the enablement pattern I have now repeated at three organizations.
A 5-page Power BI dashboard over three decades of order history, broken down by country, region, and software module to expose demand patterns that were previously invisible, plus a live Excel master tracker the team updates itself. Both are in active departmental use, and the historical patterns now connect to forward-looking demand signals such as planned data centers and desalination plants.
These four builds are the AI half of the picture. The strategy, go-to-market, and operator work behind them, and the 10+ years of business building underneath, sit on the other two pages.
Working on something where AI, product, or go-to-market meet? I’d be happy to chat.
Include: role/team + what you want solved.