Philippe
Guillamo
I govern the data inside HR systems, and I build the software that keeps it clean.
Sectors
Luxury · Retail banking · Insurance · Public transport · Cinema · Social housing
01 · Data governance
Much of the work is unglamorous: duplicate identities, ownership of reference data, rules that still hold when twenty entities share one instance. I treat HR data as a system with owners and consequences, not as a by-product of reporting.
Concretely: seven-level duplicate detection across 410,000 records, monthly cleansing campaigns built on 40 business rules, per-entity reporting, and a three-consultant remediation programme run against two immovable SAP release dates. I define who owns which data, who may change it, and what evidence survives an audit.
02 · Compliance
I built one of my tools to the high-risk requirements of the EU AI Act (Regulation EU 2024/1689): human oversight, logging, documentation. I maintain its regulatory checklist myself and keep the file audit-ready. GDPR applies end to end on the HR data domain, and Article 50 transparency applies on my own products. Compliance is not a document I commission. It is a constraint I design against.
03 · Track record
I have been on the LVMH SuccessFactors programme since 2022, first through a consultancy and, since March 2025, through my own company: core HR data, multi-country rollouts and the parts of a programme nobody wants to own. Seven countries integrated after acquisition, the United Arab Emirates among them.
Eighteen years in enterprise HR systems in all, nine of them inside a payroll software vendor, and a first career in software engineering before that.
04 · Platforms
SAP SuccessFactors Employee Central is where I work today and where I go deep: core HR design and configuration, the data model underneath, position management, payroll interfaces, and the two production releases a year.
Earlier in my career I worked on Oracle HCM Cloud and PeopleSoft. I mention them because legacy estates are real and I have lived in them, not because they are my current depth.
05 · Software & applied AI
I also build. I designed, developed and shipped a complete SaaS application alone: Next.js, PostgreSQL, Cloud Run, generative AI, eight languages, from schema to production. So I can specify an integration and then write it, and judge where generative AI earns its place in an HR process and where it does not.
On Vinca, private memories only leave their context through a controlled distillation step, gated by an adversarial bench where a judge model tries to deduce the protected secret, with a three-way vote. Five prompt regression suites run before every release: distress detection on 73 cases, false positives on 206 messages across 8 languages, adversarial attacks, distillation leak-tightness, anti-leak.
Travel Planner follows one rule: the AI writes the trip, real offers set the price. An algorithmic layer handles visas, weather and budget at zero AI cost, a solver prices candidate trips with live partner offers, and every price line says whether it is bookable, observed or estimated. About USD 0.03 per proposal.