AI platform analysis
Management could choose between two development tracks on the strength of one report, and knows which infrastructure a platform with such sensitive data belongs on.
- role
- Technical adviser
- phase
- analysis
- stack
- Next.js Supabase Spring Boot Angular Keycloak EU hosting / GDPR
What was going on
A client is building an AI-native platform that works with highly sensitive personal data. Before the next phase the team wanted to know for certain where the platform stood: is the foundation secure and scalable, does it comply with the GDPR, and what infrastructure should a platform with such sensitive data run on? I did an independent analysis: no execution, but a reasoned verdict with a concrete way forward.
What made it hard
Assessing two development tracks side by side: a feature-rich first version on Next.js and Supabase, and a younger target stack on Spring Boot, Angular and Keycloak
Testing the separation between clients and the data protection in a sector where data sensitivity is at its highest
Honest advice on EU data sovereignty: what "sovereign" really means, and which hosting fits it
Checking every finding against the live environment, so the report rests on facts and not on assumptions
How I approached it
Three weeks, read access only. First the whole platform mapped out: both codebases, the data model, the login layer and the infrastructure. Then every finding weighed on risk and impact, and checked against the live environment. The result is one report with a clear reading order: the outline for management, the reasoning for the builders, closed with a session where every question could be asked.
One report that backed the strategic choice between two development tracks
A 48-page report: architecture, security, scalability and GDPR, every finding weighed on priority
A migration plan to EU-sovereign hosting, with an honest weighing of the options
A roadmap on one A4: from today to the first customer and beyond
Follow-up work started on the basis of the findings
Need something like this?
One route: analysis, building and introducing, and maintenance if you want it. It starts with a thirty-minute intake.