AI Solutions, Business Automation & Cloud Platforms

Challenge — A German automotive OEM's material disposition team manually triaged supply-chain exceptions across SAP data — slow, reactive, and hard to scale.

Solution — A multi-agent system (planning → cognition → action layers) on AWS Bedrock that detects exceptions, runs inventory-forecast and transport analysis, and drafts escalations automatically.

Result — Disponenten get analysis and recommended actions instead of raw alerts — exception handling shifts from reactive to proactive.

Challenge — An EU insurer's staff spent hours searching scattered policy documents and internal knowledge for answers.

Solution — A retrieval-augmented chat assistant combining vector search, a knowledge graph, and guardrails — grounded answers with citations, served through a clean chat UI.

Result — Employees get cited, document-grounded answers in seconds instead of manual search across systems.

Challenge — Organizations want a trustworthy internal AI — one that answers from their own knowledge, refuses to hallucinate, and leaves an audit trail for compliance.

Solution — A full production architecture: RAG core, LangGraph agent orchestration, input/output guardrails, audit logging, and a React chat UI — deployable to Google Cloud Run via Terraform, with versioned prompts and observability.

Result — A reference-grade assistant: grounded, governed, and observable — the blueprint I deploy for regulated and enterprise clients.

Challenge — Manual handling of invoices and regulatory e-invoice formats created slow, error-prone back-office work.

Solution — A full-stack e-invoicing app with OCR-driven extraction, validation, and structured export — front end and back end deployed on Google Cloud.

Result — Invoice data is captured and validated automatically, cutting manual entry and compliance overhead.