Product
AI Product Engineering
A working prompt is perhaps a tenth of an AI product. The rest is the software: who can see what, who pays for what, what happens when a request costs more than expected. We build the entire thing, and we build the metering in from the start rather than discovering the bill later.
- Next.js
- TypeScript
- PostgreSQL
- Drizzle
- Stripe
- Better Auth
What this includes
Each of these exists because leaving it out is how these systems fail in production.
Usage and cost metering
Estimate token cost before the call, reconcile against actual usage after the stream completes, and hold it against a credit balance. You know what each customer costs you before your invoice tells you.
Multi-tenancy and permissions
Organisations, roles, seats, and per-tenant data isolation — so an AI feature cannot become a data-leak surface between customers.
Billing that matches usage
Subscriptions, per-seat, one-time, and consumption-based pricing on Stripe, wired to the metering rather than bolted alongside it.
Model configuration as product surface
Which models a given tier may use becomes a config decision, not a code change. Cheap models for cheap work, frontier models where accuracy pays for itself.
Pre + post
Cost metered before and after each call
Per-tier
Model allowlists as a billing surface
5M+
Records served in one product data layer
Where we have done this
Next step
Tell us what you’re trying to build
Most engagements start with a fixed-price audit, so the first thing you buy is a decision rather than a commitment.