Full-Stack Developer - Travel Agency SaaS (with AI)
Presupuesto: -
HOURLY / PART_TIME
⭐ 5.00 (2)
Ethiopia
api-development, react-js, node.js, react-native, next.js, mongodb, postgresql, restful-api, flutter, dart, mobile-app-development
We are looking for one strong developer to co-build a production SaaS with us — not a
freelancer to disappear after delivery. You will work directly with the founder, with full
architecture documentation provided from day one.
About the project
Profitmate is a multi-tenant SaaS platform for travel agencies. It manages bookings,
invoicing, billing enforcement, AI-powered ticket normalization, and role-based access
for agency staff and partners.
Currently live in the Ethiopian market, with the roadmap targeting multi-region
expansion — multi-currency, multi-timezone, and broader language support for ticket
normalization. You will help architect and build that expansion alongside the founder.
Stack
Backend: Node.js · Express · PostgreSQL · Prisma / Sequelize
Frontend: React · Vite · Tailwind CSS
AI: Gemini API · OpenAI API
Auth & Architecture: JWT · Multi-tenant
What you will build
• Multi-tenant auth system with 4 roles: Super Admin, Agency Admin, Partner,
Accountant
• AI ticket normalization pipeline — raw text in, structured PostgreSQL record out
— with validation and retry logic across Gemini and OpenAI
• Billing engine: invoice generation, billing enforcement middleware, usage
tracking
• Full React frontend with role-scoped dashboards, ticket management, and
reporting
• File upload handling for receipts, scoped per agency
• Multi-region architecture: currency, timezone, and localization decisions to be
made collaboratively
The AI normalization piece (core feature)
Travel agents paste raw ticket text — unstructured, inconsistent, sometimes messy. The
system must parse it into a clean structured record using an LLM, validate the output
against a schema, and retry with fallback logic if the model returns garbage.
This is not a wrapper around ChatGPT. You need to design the prompt, the
validation layer, the retry/fallback strategy, and the error handling. Experience
with LLM output reliability in production is required.
Who we need
You are a fit if Not a fit if
You have shipped AI features in production,
not just demos
You have only used AI APIs in side projects
You are comfortable migrating from
MongoDB to PostgreSQL
You need every task fully specified before
starting
You write tests before calling something done You treat PostgreSQL like a new concept
You can read existing code and make
surgical changes
You skip code review and testing
You ask hard questions before building You disappear after delivery
What you get
• Complete architecture documentation and task tracker from day one
• Direct access to the founder — no PMs, no middlemen
• A real product with real users in a market with no dominant competitor
• Flexible hours — we care about output, not when you work
How to apply
Send a short message covering three things:
• One production AI feature you have shipped — what it did, how you handled LLM
unreliability
• Your experience with PostgreSQL and any ORM you have used
• Your availability and rough rate
Abrir en Upwork