AI integration platform
Vendor-agnostic, declarative configuration standardizes AI access patterns across teams and decouples platform logic from individual model providers.
Back
Career · Internship
AI Systems Engineer · Internship
Led the design and implementation of a scalable AI integration platform at Belsimpel | Gomibo, automating workflows and adding intelligent features across the SaaS and B2B products.
Details
- Company
- Belsimpel | Gomibo
- Location
- Groningen, Netherlands · Hybrid
- Period
- May 2025 – Sep 2025
- Focus
- AI Gateway, MCP, Observability
What I did
Production-ready AI access across 30+ microservices.
Architected and launched an AI integration platform that connects with around 30 internal REST microservices through a DSL-driven AI Gateway. The gateway enforces a flexible policy framework, runs safety guardrails, and validates LLM output through schema-based parsing before it leaves the boundary.
On top of the gateway, MCP-powered automations turn LLM intent into reliable, auditable system actions. Generative pipelines for translation and product description generation cut end-to-end product publishing time across SaaS channels by roughly eight times, with G-Eval benchmarks tracking a 42% improvement in factual consistency.
Beyond the platform, I authored the adoption roadmap for Belsimpel | Gomibo covering security, monitoring and observability, rollout phases, and cost governance, plus failure-aware AI execution flows wired into Grafana, metrics, and structured logging.
Adoption roadmap
Security, monitoring, observability, rollout phases, and cost governance, mapped out so the platform could land safely across teams.
Failure-aware execution
End-to-end observability with Grafana, metrics, and structured logging keeps AI execution flows debuggable and accountable in production.