Case studies
An AI relationship-coaching platform in the US
Embedded with the founders to build an AI product from the ground up.
US, client anonymised
Context
An early-stage AI relationship-coaching platform. The founders had a vision and no product.
The gap
There was no product yet, and its core, advice on sensitive topics, could not trust an LLM's output without checks.
How we worked
We worked as the founders' embedded engineering team, directly with them and with early users, turning needs into working features.
What we built
- The MVP from the ground up on Next.js, Supabase and NestJS: magic-link sign-in, a structured assessment, free and paid report tiers and partner invitations
- A multi-dimension relationship assessment framework
- The report-generation pipeline, with evaluation and QA checks against transcripts, because sensitive advice can't trust the LLM blindly
- Retrieval and orchestration on Langfuse, Qdrant and OpenRouter, and the product's core AI agent
- Automated user reports through n8n and Zapier
- A retention loop of debriefs, regular check-ins and weekly reviews
- Onboarding flows, UX copy, email templates and user research
The loop
- Find. Worked directly with the founders and early users to turn their needs into features.
- Build. Designed the assessment framework and the architecture for retrieval, orchestration and evaluation.
- Ship. Shipped the MVP to early users, with free and paid report tiers.
- Adopt. Built onboarding, automated reports and the retention loop, and ran user research.
Outcome
Outcome figures are being confirmed with the client before we publish them.
Stack
Next.js, Supabase, NestJS, Langfuse, Qdrant, OpenRouter, n8n, Zapier
Let's find where AI fits in your business.
Book a discovery call. We'll tell you where AI would pay off in your systems, and what the first build would take.