I work where enterprise systems meet modern AI — the seam where a demo becomes a process, and a process has to survive an audit, a release calendar, and a business user who does not care about your architecture.

Practically, that means three things.

Agents that ship

Designing and building agentic systems inside real organisations: tool contracts, permissions, human gates, traces, evaluation sets, and a cost model someone signs off. The interesting engineering is never the prompt — it is the blast radius.

S/4HANA programmes

Conversion and implementation work on the SAP side: custom code triage, clean-core governance, data and cutover readiness. I have seen enough programmes to know that the plan is rarely the problem and the definition of done usually is.

Machines that see

Outside the day job: a 5-inch reconnaissance drone built from parts, a Raspberry Pi 5 running PyTorch on board, and the geometry that turns a bounding box into a coordinate. The build log is here, including the parts that failed.

How I work

  • Constraint first. I ask what is actually blocking you before proposing an architecture.
  • Small, verifiable steps. A working slice in a week beats a design document in a month.
  • Written artefacts. Decisions get written down with an owner and a date, or they get re-litigated.
  • Honest about uncertainty. If I have not run it myself, I say so.

What I take on

  • Agent architecture reviews — existing pilot, no measurable value, unclear ownership.
  • S/4HANA programme rescue — custom code, clean core, test and cutover readiness.
  • Embedded vision prototypes — from "we think this is possible" to a flying demonstration.
  • Writing and technical content — the pages on this site are the portfolio.

Contact: contact@nourredine.tech. Tell me the constraint you are stuck on, not the title of the project — it makes the first reply useful.