ABOUT THE PROJECT
WFD (What's for Dinner) is a venture-backed food discovery platform launching in Miami, Florida. We are building the first conversational AI food discovery app where an AI agent named Joy acts as your personal foodie friend.
Instead of scrolling through hundreds of restaurants on Uber Eats or DoorDash, users simply tell Joy what they're in the mood for — by voice or text — and Joy recommends the perfect local restaurant instantly.
Our tagline: "High Tech, High Touch. Hospitality at Your Fingertips."
This is a Phase 1 MVP build for investor demos, pilot restaurant onboarding, and early user validation. We are looking for a talented developer who wants to be part of something genuinely exciting from the ground floor. Strong performance on this project opens the door to a long-term development partnership as we raise seed funding and scale nationally.
WHAT WE ARE BUILDING
A production-quality web application (mobile-first, WebView compatible) with the following components:
1. Joy — Conversational AI Agent
Joy is the heart of the product. She must feel warm, human, and knowledgeable — like a foodie friend, never a generic chatbot.
Conversation capabilities:
Understands natural language food queries ("I'm sick and want something warm and comforting" / "spicy food under $15 near me" / "something good for a date night")
Responds with confident, personalized restaurant recommendations
Remembers the user's name and greets them personally ("Welcome back, Mike")
Handles context — time of day, mood, dietary preferences, price range
Upsells naturally and hospitably ("A lot of people who order from there also love their...")
Bilingual — English and Spanish, auto-detected from user input with manual toggle
Escalates gracefully when uncertain ("Let me think about that one for you...")
Powered by: OpenAI GPT-4o (we provide API key)
2. Voice Input + Joy Voice Output
Voice Input (Speech to Text):
User taps a microphone button and speaks their request
Real-time transcription displayed as they speak
Powered by: AssemblyAI streaming STT (we provide API key)
Must work reliably on mobile (iOS Safari + Android Chrome)
Graceful fallback to text input if microphone is unavailable
Joy Voice Output (Text to Speech):
Joy speaks her responses out loud in a warm, human voice
Powered by: ElevenLabs TTS (we provide API key and purchased voice)
Audio plays automatically after Joy responds
User can tap to replay or skip
Must not feel robotic — voice quality is critical to the product experience
3. Restaurant Recommendation Engine
Data architecture:
Restaurant database stored in Supabase (PostgreSQL)
We will provide initial seed data for 20-30 Miami restaurants
Each restaurant record includes: name, cuisine type, price range, neighborhood, description, keywords, website URL, Google Maps link, phone number, hours
Recommendation logic:
Joy matches user query to restaurants using a combination of OpenAI semantic understanding and keyword matching against restaurant database
Ranking factors: relevance to query, cuisine match, price range match, neighborhood proximity (based on user-provided location or zip code)
Returns top 1-3 recommendations per query
Each recommendation includes: restaurant name, Joy's personalized description of why she's recommending it, cuisine, price range, and action buttons
Action buttons per recommendation:
"Get Directions" (Google Maps link)
"Order Online" (restaurant website or delivery platform link)
"Tell Me More" (Joy provides more detail about the restaurant)
4. User Experience & Interface
Design philosophy: Minimalist, premium, mobile-first. Think Apple-level simplicity meets warm hospitality. No clutter. No overwhelm.
Key screens:
Welcome / name capture screen (first visit only)
Main chat interface with Joy
Recommendation cards (clean, visual, actionable)
Language toggle (EN / ES)
Branding:
App name: WFD
Agent name: Joy
Primary color: Dark navy #1B3A5C
Accent color: Warm orange #E87722
Font: Clean sans-serif (Inter or similar)
Joy should have a subtle animated presence (pulsing mic indicator, typing indicator)
Device compatibility:
Mobile-first (iPhone and Android)
Must work inside a WebView (for future restaurant app embedding)
Responsive desktop view
Offline-graceful (clear error states if no connection)
5. User Memory & Personalization
User name stored locally and in database on first visit
Session history stored per user (last 5 queries and recommendations)
Joy references past interactions naturally ("Last time you ordered Thai — want something different tonight?")
Language preference remembered across sessions
Basic preference learning (if user frequently asks for healthy food, Joy adjusts recommendations accordingly over time)
Storage: Supabase for user profiles and session history
6. Basic Admin Panel
A simple password-protected admin interface where the WFD team can:
Add, edit, and remove restaurants from the database
View basic analytics (total queries, most recommended restaurants, most popular keywords)
No complex dashboard required — functional is fine
7. Analytics & Event Tracking
Track the following events (store in Supabase):
Session started
Query submitted
Language used (EN/ES)
Recommendation shown
"Order Online" clicked (this is the key conversion metric)
"Get Directions" clicked
"Tell Me More" clicked
This data is critical for proving value to both investors and restaurant partners.
8. Deployment & Infrastructure
Frontend deployed to Vercel
Backend/database on Supabase
Environment variables properly configured
Staging environment + production environment
Basic rate limiting to prevent API abuse
Clean GitHub repository with README and setup instructions
TECH STACK
14+ with TypeScriptStylingTailwind CSSAI/LLMOpenAI GPT-4o APISpeech to TextAssemblyAI StreamingText to SpeechElevenLabs APIDatabaseSupabase (PostgreSQL)Auth (admin)Supabase AuthHostingVercelVersion ControlGitHub
DELIVERABLES
Fully working web application deployed to live Vercel URL
Staging environment for testing
Clean GitHub repository with all source code
Basic admin panel (password protected)
Supabase database with seed data for 20-30 Miami restaurants
README with local setup instructions and environment variable documentation
Brief handover call or video walkthrough of the codebase
WHAT THIS PROJECT DOES NOT INCLUDE
No payment processing or in-app ordering (users are redirected to complete orders elsewhere)
No native iOS or Android app (web app only in Phase 1)
No complex loyalty or rewards system
No POS integrations
No multi-city support in Phase 1 (Miami only)
TIMELINE
4-6 weeks from project start
We understand quality takes time. We would rather receive something excellent in 6 weeks than something rushed in 2 weeks.
IDEAL CANDIDATE
You are a strong fit for this project if:
You have built AI-powered or GPT-integrated web applications before (please share links)
You have experience with voice/audio in web apps (STT or TTS integration)
You are comfortable with, TypeScript, and Supabase
You write clean, well-structured code you'd be proud to show
You communicate proactively and flag issues early
You care about UI/UX quality, not just functionality
You are excited about the product and want to be part of something growing
SKILLS NEEDED
, TypeScript, React, OpenAI API, AssemblyAI, ElevenLabs, Supabase, PostgreSQL, Tailwind CSS, WebSocket, Audio Streaming, Vercel, UI/UX Design, REST API Development
Contract duration of 1 to 3 months.
Mandatory skills: Fastify, API Development, PostgreSQL Programming, Automatic Speech Recognition, Voice API, Real-time Applications,, TypeScript, Redis