Named where the client cleared it, anonymised where they did not. Each of these follows the same structure, including what went wrong, which is usually the part worth reading.
11 years in production without a rating rewrite
A rating and charging system built when there was nothing on the market to buy, which then had to survive a decade of tariff changes, a new regulatory settlement model, and a company that grew around it.
Several hundred agents, four months to beta
An agent desk and the middleware under it, built against a young contact-centre platform whose API was still growing while the project ran, where the schedule risk sat in another company’s backlog.
Nine years in production, audience up 25%, two services extracted
A CRM for commercial property, kept in production for nine years while its backend platform, its frontend generation and its database were all replaced underneath it. The interesting decision was how little of it became services.
3–5 fps to a sustained 15–20 on the same hardware
A TensorFlow Lite and OpenCV pipeline on cheap ARM boards guarding a residential settlement. The model was never the constraint. Frame-time spent outside inference was, and then the physical world was.
Signal worth 0.5–1.5 bp against 11 bp of cost. No live order ever sent.
A six-layer crypto execution platform I built for myself, and the two weeks of measurement that proved the strategy on top of it could not pay for its own fees. The engineering was fine. The economics were never checked, until they were.
If it isn't my kind of problem, I'll say so and point you somewhere better. Direct email works too: hi@realgeek.biz.