AILive

Propellus

AI-powered travel platform that personalizes your journey

#Next.js
#Nest.js
#PostgreSQL
#AWS
Propellus: AI-powered travel platform that personalizes your journey, built by Zeeshan Ashraf

#problem

Propellus is an AI travel platform that curates personalized journeys for each user, moving beyond standard search-and-book flows toward tailored recommendations. It addresses the problem of generic, one-size-fits-all travel planning by structuring user preferences and trip data so that itineraries can be adapted to the individual. The product spans a user-facing web application and a backend service that manages personalization, persistence, and the AI-driven recommendation logic.

#what-i-built

01

I architected the full Propellus platform end to end, owning the Next.js frontend, the Nest.js backend, the PostgreSQL data layer, and the AWS infrastructure.

02

I built the Next.js frontend that delivers the personalized travel experience and trip-planning interface to users.

03

I designed and implemented the Nest.js backend services and REST APIs that drive the AI personalization and recommendation logic.

04

I modeled the PostgreSQL schema for user preferences, trips, and itinerary data that underpins personalized journeys.

05

I deployed and operated the platform on AWS, setting up the cloud infrastructure and CI/CD pipeline for releases.

#stack

#Next.js
#Nest.js
#PostgreSQL
#AWS

#outcome

Production AI travel platform live at propellus.co: end-to-end ownership across Next.js frontend, Nest.js backend, PostgreSQL data layer, and AWS infrastructure.

#key-decisions

01

Chose Nest.js for its modular, TypeScript-first architecture, keeping the AI personalization service cleanly separable from the API routing layer and independently testable.

02

Used PostgreSQL over a NoSQL store to model user preferences and itinerary data relationally, giving the personalization engine flexible querying without data consistency trade-offs.

03

Deployed on AWS with containerized services to support horizontal scaling as user volume grows rather than a fragile single-instance setup.

04

Isolated AI personalization logic in a dedicated backend service (not in the frontend) so future LLM provider swaps require no UI changes.

Building something similar?

Let's talk about your project. No obligation.