AI product engineering in Groningen, shipped to production.
RAG pipelines, agents, embeddings, evaluation harnesses. We ship AI products that work in production, not demos that fall over on the first real user. AI-native delivery, production-grade output, shipped from Groningen in your timezone.
What we build
RAG that survives real traffic
Chunking strategy, hybrid retrieval, reranking, citation traces. pgvector or Pinecone, your call, with eval harness from day one.
Agents with guardrails
Tool calls, structured outputs, retry budgets, deterministic test paths. We've shipped agents that run unattended in production with audit trails.
Streaming UI patterns
Token-by-token rendering with Vercel AI SDK, cancellable streams, optimistic UI. The patterns visitors expect from ChatGPT, adapted to your domain.
Cost + latency observability
Per-request token tracking, latency dashboards, A/B between models. We treat AI like any other production system: instrumented or it doesn't ship.
Where this fits
You're a Groningen-based scale-up adding AI features and your team has shipped React, not LLM apps.
You're a RUG spin-out with an ML thesis and need to turn the model into a product visitors actually use.
You're a regional B2B operator and an AI assistant on your internal data would replace 80% of the back-office back-and-forth.
Tech stack
- TypeScript
- OpenAI / Anthropic
- RAG pipelines
- pgvector
- Vercel AI SDK
- BullMQ
Want this for your team?
30 minutes with a founder or senior engineer. We'll scope what you need and tell you straight whether Stacklane fits.
Book a Free CallRelated capabilities