The problem
Travel discovery is not one API call. An open request such as “somewhere warm, four nights, flexible dates, under £500” creates a huge candidate space before supplier shopping even begins.
Case study · MyEscapePlan
From open-ended travel intent to production search.
MyEscapePlan is a production travel-intelligence platform built from zero across AI decisioning, data pipelines, consumer and advisor applications, APIs, SDKs and third-party travel integrations. The work moved directly from customer and partner requirements to production implementation.
LIVE SYSTEM
Intentwarm · 4 nights · flexible · budget
Decisioningconstraints → candidates → rank → shop
ProductionAPI · database · workers · suppliers · UI
The engineering
The system turns natural-language intent into structured constraints, generates candidate destinations and date variants, ranks them, controls fan-out, calls live inventory providers and explains the result back to the user.
The production work
FastAPI services, PostgreSQL, Celery background workers, Redis caching, supplier APIs, SDKs, observability, cloud infrastructure, CI/CD, and consumer and advisor applications all have to behave as one production system.
Why it matters here
That is forward deployed engineering in practice: start from real customer and partner requirements, work across product and technical boundaries, measure quality and cost, and carry the deployment through to production.