HK Smart Explorer: AI Tourism App

Inspiration

Driven by Hong Kong’s 2025 tourism struggles: 49M arrivals vs. 65M pre-2019, 6% spending drop, overcrowding (51.5M in Aug), and competition from Singapore/Tokyo.
HKTB app’s issues—fragmented info, clunky interface (3.5/5 ratings), no personalization (e.g., rainy hikes)—sparked our vision for an AI-powered, user-centric solution to boost satisfaction and recovery.

What it does

Enhances HKTB’s app with AI for personalized, budget-friendly, crowd-free travel.
Features TSFM for weather-based timing (e.g., “Victoria Peak in dry Nov”), trip planner (HK$800–3,000/day), budget mode, spending forecasts, auto-blogs, feedback system, HKTB analytics, vendor recs, and monastery audio.
Uses Amazon Q Developer for itineraries and Kiro for predictions/content.

How we built it

Python stack (Prophet, scikit-learn, Streamlit) with AWS SageMaker, QuickSight, Polly.
Trained TSFM on HKO weather (1884–2025); used HKTB stats, Google APIs for crowds/deals.
Amazon Q Developer generates itineraries (e.g., “5-day eco-trip, HK$5,000”); Kiro powers climate, sentiment, blogs.
Privacy via opt-in data; hosted on GitHub.

Challenges we ran into

Limited real-time crowd APIs for Hong Kong hotspots.
Balancing personalization with crowd delegation via K-means clustering.
Ensuring offline access for budget travelers.
Tuning Amazon Q Developer/Kiro for clear, culturally nuanced prompts.
Preprocessing diverse weather data (typhoons, humidity) for TSFM.

Accomplishments we’re proud of

Accurate TSFM for weather-optimal visits, preventing disappointments.
Budget planner for low-cost trips (HK$800/day), countering 6% spending drop.
Delegated vendor recs boosting 15% festive sales while spreading crowds.
HKTB dashboard for real-time planning (e.g., “5,000 to Peak”).
Seamless Amazon Q Developer/Kiro integration in a demo-ready app.

What we learned

Combining Amazon Q Developer (content) and Kiro (predictions) creates intuitive experiences.
Robust preprocessing critical for Hong Kong’s weather/crowd data.
Privacy (opt-in, anonymized) builds user trust.
Cultural nuance in AI prompts (e.g., monastery histories) enhances engagement.

What’s next for HK Smart Explorer

Add offline caching for recommendations/maps.
Expand crowd data via X API.
Integrate AR for monastery tours with Amazon Q Developer.
Partner with vendors for deals (e.g., 10% off festive dishes).
Scale with AWS Lambda for 2026’s 60M+ visitors.

Built With

• amazon-developer
• artificial-intelligence
• kiro
• machine-learning
• python