fas7ny فسحني
Mohnd Bata
Graduation Project · 2025 Production Ready
fas7ny فسحني
AI-Powered Intelligent Travel Planner
A full-stack mobile application fusing BERT-based NLP, Genetic Algorithms, PDDL planning, and A* pathfinding into a unified AI pipeline — generating hyper-personalized travel itineraries with real-time pricing, smart recommendations, and seamless payment processing.
Flutter Node.js MongoDB Firebase BERT Genetic Algorithms PDDL A* Redis JWT
Demo Video
BERT
NLP Understanding
GeneticAlgorithm
Route Optimization
PDDL
AI Trip Planning
A*
Smart Pathfinding
10+
Core Modules
2025
BUE Graduation
Problem
Travel planning is broken.
Modern travelers face a fragmented, time-consuming experience. Planning a trip requires juggling multiple apps — search engines, review sites, booking platforms, and payment gateways — none of which understand the traveler's actual preferences.
Generic recommendation engines suggest popular destinations with no regard for budget constraints, group dynamics, time availability, or personal interests. The result is either decision paralysis or a mediocre trip that could have been much better.
The core challenge: How do you build a system that truly understands what someone wants, reasons about their constraints, and generates an optimal, executable travel plan — all in real-time?
Pain Points
Fragmented booking across 5+ platforms
Generic recommendations ignoring context
No budget or time optimization
Manual itinerary building takes hours
No support for natural language input
Solution
One AI pipeline. Complete travel intelligence.
Fas7ny replaces all those fragmented tools with a single intelligent platform. A user types their request in natural Arabic or English — "Plan a 3-day Cairo trip for two people on a 2000 EGP budget" — and the AI engine does the rest.
The system processes the intent through BERT NLP, builds a personalized user profile, then passes the problem to a Genetic Algorithm to find the globally optimal set of destinations. A PDDL planner sequences the activities into an executable schedule, while A* search computes the most efficient travel routes between locations.
The result is a complete, personalized, budget-respecting itinerary — delivered in seconds, with real-time pricing and integrated payment.
What Fas7ny Does
Natural language input in Arabic & English
Budget-aware, preference-personalized trips
Globally optimal routes via Genetic Algorithms
PDDL-planned, A*-refined itineraries
Real-time pricing & payment in one app
Architecture
System Design
Client
Flutter App
Mobile UI · State Mgmt
Backend
Node.js API
REST · Express · JWT
Firebase Auth
OAuth · Sessions
AI Engine
BERT NLP
Intent · Entities
Genetic Algorithm
Route Optimization
PDDL Planner
Itinerary Sequencing
A* Search
Optimal Pathfinding
Data
MongoDB
Primary Store
Redis
Cache · Sessions
Firebase
Realtime · Storage
AI Pipeline
How the Intelligence Works
01
Natural Language Input
User enters a free-text travel request in Arabic or English. No forms, no dropdowns — just a natural sentence like "خطط لرحلة للإسكندرية لمدة يومين بميزانية ١٥٠٠ جنيه".
Arabic NLPFree Text
02
BERT NLP Processing
A fine-tuned BERT model extracts intent, entities (destination, duration, budget, group size, preferences), and sentiment. Handles both Modern Standard Arabic and Egyptian dialect.
BERTNERIntent Classification
03
Recommendation Engine
Based on extracted profile and user history, a collaborative filtering model suggests candidate places, experiences, and activities that match the user's profile and budget constraints.
Collaborative FilteringUser Profiling
04
Genetic Algorithm Optimization
A GA evolves a population of candidate itineraries over multiple generations, optimizing for: satisfaction score, total cost, travel time, and constraint satisfaction. Finds near-optimal solutions in polynomial time.
Genetic AlgorithmMulti-objective Optimization
05
PDDL Trip Planning
PDDL (Planning Domain Definition Language) structures the selected activities into a temporally consistent, action-sequence plan. Handles dependencies between activities (e.g., hotel check-in before museum visits).
PDDLTemporal PlanningConstraint Satisfaction
06
A* Pathfinding & Final Itinerary
A* search computes the most efficient travel routes between locations using real-world distance and traffic heuristics. The final itinerary is packaged with real-time pricing, integrated booking, and payment processing.
A* SearchDynamic PricingPayment
Modules
Core Application Modules
NLP Understanding
Fine-tuned BERT model for Arabic/English intent classification, named entity recognition (destinations, dates, budget), and sentiment analysis of user preferences.
BERTTransformersArabic NLP
Recommendation Engine
Collaborative and content-based filtering to surface contextually relevant destinations, activities, and restaurants. Learns from user interactions and continuously improves suggestions.
Collaborative FilteringUser ProfilesRedis
Optimization Engine
Multi-objective Genetic Algorithm that balances cost, satisfaction, and travel time simultaneously. Evolves itinerary populations through selection, crossover, and mutation to find optimal solutions.
Genetic AlgorithmsMulti-objectiveHeuristics
AI Trip Planning
PDDL-based formal planner that converts the optimized destination list into a structured, time-consistent daily schedule. Handles temporal dependencies and activity sequencing constraints.
PDDLTemporal LogicAI Planning
Dynamic Pricing
Real-time price aggregation and comparison for hotels, activities, and restaurants. Tracks price fluctuations, applies budget constraints, and suggests alternatives when costs exceed limits.
Real-time APIsBudget EngineNode.js
Payment Integration
Secure, in-app payment processing for bookings. Integrates with payment gateways, manages transaction state, handles refunds, and generates booking confirmations — all within the Flutter app.
Payment GatewayJWTFlutter
Technology
Full Tech Stack
Mobile
FlutterCross-platform UI
DartApplication Language
ProviderState Management
Backend
Node.jsRuntime
Express.jsREST API
JWTAuthentication
Databases
MongoDBPrimary Database
FirebaseRealtime & Auth
RedisCaching Layer
AI & ML
BERTNLP Model
Genetic AlgorithmsOptimization
PDDLAI Planning
A* SearchPathfinding
PythonAI Services
Want to see more?
Like this project
