From Marton Csernai’s experience

A multi-agent AI coaching system with auditable, GDPR-compliant traces

For a wellness startup, Márton architected a 4-agent LLM coaching system with dynamic 6-layer prompt assembly and emotional intensity mapping, with auditable GDPR-compliant traces, and built the mobile app.

Industry
Health and life sciences
Marton’s role
Architect and developer

The business need

A wellness startup needed an AI coaching system and a mobile app to deliver it. The system had to respond to the emotional intensity of a conversation and keep auditable, GDPR-compliant traces of what it did.

What Marton did

  1. Architected a coaching system built from four LLM agents.
  2. Designed dynamic prompt assembly in six layers.
  3. Added emotional intensity mapping to the system.
  4. Kept fully auditable, GDPR-compliant traces of the system's work.
  5. Built the accompanying mobile app.

Technology and approach

  • Four LLM agents
  • Dynamic 6-layer prompt assembly
  • Emotional intensity mapping
  • Auditable, GDPR-compliant traces
  • Mobile app

Decisions this experience speaks to

  • We handle sensitive health or financial data - how do we build AI that stays GDPR-compliant?
  • How do we make our AI auditable and explainable to customers and regulators?
  • How do we move from simple prompts to RAG and agentic workflows that handle our proprietary logic?

Related services and use cases

Sources

  1. Marton Csernai’s profile on CTO Grand Rounds