All journeys

I want to build AI without giving away my data

Privacy is an architecture you can point at, not a promise in a settings screen. This route keeps the useful work local, measures the trade-offs, and ends with a system whose data boundary another person can verify.

You end with: A useful system running locally, with its quality, latency, energy use and data boundary measured rather than assumed.

Explore

Local does not mean free. Measure the power and carbon cost before choosing hardware or claiming that keeping data close is automatically the better system.

  1. Where the energy actually goes — Joules, not vibes. Measure what your own work costs before joining an argument about data centres. (3 lessons, 65 min)

Build

Start with a small workload, write the tests before moving it, then measure what local inference changes in quality, latency and control.

  1. Coffee break: the summariser you can prove is wrong — Twenty minutes to build it, twenty more to find out where it fails. The second half is the point. (2 lessons, 40 min)
  2. The eval you actually need — Stop reading leaderboards. Build the harness that measures your task, on your inputs. (4 lessons, 8 min)
  3. Local or nothing — Run open-weight models on hardware you own, and know what they cost you in watts and seconds. (5 lessons, 6 min)

Launch

A private prototype is still only a prototype. Run it on real inputs with permission, an explicit boundary, a rollback and evidence it helped.

  1. Ship one automation — Take one real process end to end, put it in front of people, and measure what it did for six weeks. (26 lessons, 1255 min)