welcome
our slogan
hunger value
our values
app flow
app flow demo

Haaangry

Inspiration

We are hyped around the idea of emerging AI-Centered Companies. That was the main inspirations we had along these 40 hours.

Y Combinator "AI-First" company challenge inspired us to reimagine food ordering. We looked at DoorDash's end-to-end experience and asked: what if AI could predict what you want to eat before you even know it? We wanted to combine AI-powered personalization, data-driven UX, and conversational ordering into a single seamless experience where discovery and ordering happen in the same place.

What it does

Haaangry is a TikTok-style food discovery and ordering app powered by Claude AI.

Vertical video feed of cooking content with gesture-based navigation
Swipe left → 6 AI-curated recipes (3 articles + 3 videos)
Swipe right → 3 AI-matched restaurants with 3 recommended dishes each
Clickable menu links that open real restaurant websites
In-app ordering with cart and checkout
Voice2Order and TextOrder AI assistants for ordering
User profiles that track preferences, cravings, and order history over time

Claude analyzes each video's title, description, and tags to recommend contextually relevant restaurants and dishes in real-time.

How we built it

Backend

FastAPI server with hot-reload and production-ready configuration
Claude AI integration with strict JSON schema enforcement
Parallel processing via ThreadPoolExecutor for recipe searches
Pydantic 2.x for data validation and type safety
CORS configured for iOS client requests
Static file serving for locally hosted video assets
noSQL-based storage: 603-line restaurants.json (20 restaurants × 20 menu items each)

Frontend

SwiftUI + MVVM architecture for clean separation of concerns
AVFoundation for smooth 60fps video playback
Async/Await networking with type-safe Codable models
Gesture recognition for natural swipe navigation
PlayerPool optimization to reuse AVPlayer instances and minimize memory
Sheet-based modals for recommendations and recipes
Clickable menu links using SwiftUI Link component
Error recovery with retry mechanisms and graceful fallbacks

Integration

REST API (JSON over HTTP) between iOS app and FastAPI backend
ThreadPoolExecutor for parallel Claude API calls (article + video recipe searches)
End-to-end schema alignment ensuring menu URLs flow from backend to UI
Graceful fallbacks on JSON parse errors or network issues

Challenges we ran into

Schema drift between Python (Pydantic) and Swift (Codable) models
Duplicate struct definitions causing compilation errors
SwiftUI View Builder constraints in computed properties
Claude API rate limits during parallel requests
JSON validation across a 603-line restaurant dataset
Memory optimization for multiple concurrent video players
Merge conflicts while coordinating parallel development
Balancing technical sophistication with hackathon time constraints

Accomplishments that we're proud of

App that feels like something fundamentally novel
Zero compilation errors in final codebase
100% menu URL coverage across all 20 restaurants
Expanded catalog from 10 → 20 restaurants (400 total menu items)
Sub-3-second average AI recommendation latency
60fps video playback on device
End-to-end clickable menu links from backend to UI
Production-ready backend running on port 8000 with strict validation
10 cuisine types (exceeded 8+ target)

What we learned

Product

Balancing AI complexity with intuitive user experience
Importance of gesture-based mobile navigation
Value of immediate menu access via external links
Trade-offs between feature breadth and implementation depth

Technical

FastAPI architecture and hot-reload workflow
Claude AI JSON schema enforcement techniques
AVFoundation video lifecycle management
Pydantic validation and cross-platform data model alignment
SwiftUI MVVM composition and async/await networking
Parallel processing in Python with ThreadPoolExecutor

What's next for haaangry

Immediate Next Steps

DoorDash API integration for live order placement
Stripe / Apple Pay / Google Pay for payment processing
AWS Maps / Google Routes API for delivery tracking
Persistent order history and user accounts

Medium-Term Enhancements

Fine-tuned LLM or custom ML model for personalized recommendations
Multi-turn conversation-to-order dialogue system
User authentication with saved preferences
Delivery-time-sensitive filtering based on location
Real-time order tracking with push notifications

Long-Term Vision

Social features (share videos, reviews, ratings)
Restaurant partnerships for exclusive deals
Dietary filters (vegan, halal, allergies, kosher)
Multi-language support for global expansion
Analytics dashboard for engagement patterns

Tech Stack

Backend: Python 3.10+, FastAPI 0.120.0, Anthropic Claude 3.5, Pydantic 2.12.3, Uvicorn
Frontend: Swift 5+, SwiftUI (iOS 16+), AVFoundation, Combine
AI/ML: Anthropic Claude API for semantic matching and recipe discovery
Data: JSON storage (20 restaurants × 20 items, 226 videos)