5 deep reads

Case Studies

Long-form reads on specific engagements. What the problem actually was, what we tried, what worked, and what I would do differently. Honest about the parts that did not go well.

← Home

  1. Predictive Maintenance at Fleet Scale with Sparse Failure Data

    How a 25-person delivery team shipped a production ML platform for 6,000+ industrial companies when the thing everyone agreed was necessary (labelled failure data) barely existed.

    Industrial SaaS · 11 months

  2. Serverless IoT at 200K messages per second, for 1.6M vehicles

    A legacy telematics platform was folding under load. We rebuilt the ingest layer on serverless primitives. Cost held flat through 3x traffic growth. Here's what that took.

    Nordic telematics · 18 months

  3. Real-time ML for the Bundesliga

    The Deutsche Fußball Liga wanted broadcast-grade real-time insights during matches. We built the platform that scores match moments in under two seconds of latency, for every live broadcast.

    European media, professional sport · 14 months, two phases

  4. Scaling Serverless Expertise Across EMEA Professional Services

    AWS Professional Services EMEA had serverless knowledge concentrated in four people. Every deal needed one of them in the room. We fixed the structural problem instead of the staffing symptom.

    Consulting practice, internal · 8 months, ongoing

  5. Open-sourcing a Security Hub Solution Across Fortune 500

    An enterprise customer needed multi-account security posture management but had no product budget. The solution we built for them ended up adopted by ten or more Fortune 500 organizations across industries.

    Enterprise security, financial services · 6 months initial, 2 years adoption