Skip to content

Case studies

What I built, and what runs today

For each project I describe the problem, my contribution, its current status and my role. Clients remain anonymous; documents and materials are shared only where permitted.

AI adoption programme at a high-tech industrial company

Industry · anonymised case
Problem
Development teams that already use AI, each in their own way, without shared rules, with confidential data that cannot leave the company.
Approach
Guidelines written with the teams, a review of the internal AI platform, an on-premises AI model with per-person access, an AI Act compliance plan.
Outcome
Programme under way, with shared rules, an inventory of tools and an on-premises model configured not to send data to external services.
Role
External advisor and technical design.
Illustrative AI agent architecture: individual access and cost limits, an on-premises model, tools with least-privilege access, approval of irreversible actions, necessary event logging with defined retention, and pre-release tests. People personal access Router keys and cost limits On-premises model installed on your servers Tools and data least privilege Human approval before irreversible actions Audit trail events, access and retention Automated tests at every release

Platform for a non-profit organisation: CRM, identity, workflows and AI agents

Non-profit · proprietary platform
Problem
Data and processes scattered across several domains (contacts, events, identity, content), with no single CRM and no automation.
Approach
Architecture and development of six connected applications covering CRM scoring and workflows, shared identity, events, AI-powered analysis, race timing and the website.
Outcome
Platform in production, with continuous releases and automated checks before every release.
Role
Analysis, architecture, development, release and operations.

Being found by AI assistants

Non-profit · continuous measurement
Problem
More and more people ask an AI assistant instead of searching on a search engine: an organisation does not know whether, or how, it is cited.
Approach
Automated weekly tracking of the answers several AI assistants give to a fixed set of questions, repeated to account for the variability of the answers.
Outcome
An indicator of citation frequency across the sampled questions and assistants, used to guide content decisions. It does not measure overall online visibility or guarantee future citations.
Role
Concept, development and analysis.

How the measurement works

  1. Fixed questions the same every week
  2. Several AI assistants queried automatically
  3. Repetitions to account for variability
  4. Citation rate how often the organisation appears
  5. Decisions what to publish and what to fix

Brand reputation search and comparison engine

Technical collaboration · anonymous case study
Problem
Measure and compare brand reputation in a structured, repeatable way.
Approach
Technical development of the search and comparison engine, based on the partner’s reputation scoring method.
Outcome
Engine in operational use by the partner.
Role
Technical contribution to development.

Course platform with a private area and certificates

Training · this site
Problem
Deliver video courses with access restricted to enrolled users, progress tracking and certificates.
Approach
Platform built on Laravel and Filament: video lessons, progress tracking, student area, certificate templates.
Outcome
Running with the local public transport courses; the basis for the AI literacy programme.
Role
Analysis, integration and development.

Want to see how I would do it for you?

Tell me about a process or project. I’ll explain the feasibility, the main risks and the first step I would recommend.

Request an assessment proposal