AI-powered local meal intelligence for Africa.
Built during The Build — NSK AI / Udara Project 2026
| Member | Country | Role |
|---|---|---|
| [CREW MEMBER 1 — Name] | [Country] | [Role] |
| [CREW MEMBER 2 — Name] | [Country] | [Role] |
| [CREW MEMBER 3 — Name] | [Country] | [Role] |
GitHub: https://github.com/Logisite/Team-B
| Document | Link |
|---|---|
| Who Hurts Most | docs/who_hurts_most.md |
| Validation Conversations | docs/validation_conversations.md |
| Architecture Diagram | docs/architecture_diagram.md |
| Tech Stack Writeup | docs/tech_stack.md |
The people who hurt most are everyday Africans — starting with Nigerians — who are trying to eat better but are forced to make food decisions under real constraints.
They are:
- Students living on limited weekly budgets
- Young professionals balancing health and cost
- Parents feeding families under rising food prices
- Gym-goers trying to hit nutrition goals cheaply
- People with dietary goals but little nutrition guidance
- Low-income households trying to avoid unhealthy eating patterns
These users are not asking for perfect diets. They are asking:
- "What is the best meal I can make with what I have?"
- "What can I eat today that is healthier but still affordable?"
- "Which local food gives me the nutrients I need?"
- "What alternatives can I use if I cannot afford this ingredient?"
- "How do I eat better without eating foreign foods?"
Our initial market is Nigeria.
The first users are:
- University students
- Early-career workers
- Urban and semi-urban households
- Health-conscious young adults
Many existing nutrition apps are built around Western diets and ingredients. They suggest foods that are unavailable, culturally unfamiliar, or unrealistic for local users.
Examples:
- Zucchini
- Avocado-heavy diets
- Imported protein plans
- Expensive supplements
This creates a gap between nutrition advice and practical reality.
The pain is not just hunger.
The real pain is uncertainty.
Users struggle daily with:
- Not knowing what balanced meal to prepare
- Choosing meals under tight budgets
- Limited understanding of nutrients in local foods
- Lack of affordable healthy options
- Poor diet planning knowledge
- Conflicting health advice online
- Wasting money on inefficient food choices
Even users who want to eat healthier often fail because existing tools:
- Ignore local food culture
- Ignore local availability
- Ignore price sensitivity
- Ignore regional recipes
- Assume ideal conditions instead of reality
The result is unhealthy eating patterns, poor nutrition decisions, avoidable deficiencies, and frustration.
NutriLocal helps users make smarter food decisions using:
- Local ingredients
- Local recipes
- Budget constraints
- Nutrition goals
- Ingredient availability
- Dietary preferences
The system recommends practical meals that users can actually make.
Instead of: "Eat salmon and quinoa."
NutriLocal says: "Here are 3 affordable balanced meals you can make this week in Lagos under ₦4,000 using foods available near you."
Three trends make this the right time:
- Rising food costs across Africa
- Growing health awareness among young Africans
- AI systems becoming capable of contextual recommendation and reasoning
The internet already contains nutrition knowledge. The missing layer is local intelligence.
That is the problem NutriLocal solves.
Our goal was to validate:
- Whether users struggle with food decision-making
- Whether budget-aware nutrition is a real pain point
- Whether users trust local-first recommendations more than generic diet apps
- Whether ingredient-aware recommendations are valuable
- Whether users would repeatedly use such a system
University student, Lagos, Nigeria
The user explained that food budgeting is a weekly struggle and that most healthy food advice online feels unrealistic.
"I know noodles every day is not healthy, but most apps recommend foods I cannot afford or even find nearby."
Users want practical meal alternatives within budget constraints.
Young professional, Abuja, Nigeria
The user wanted healthier eating options but struggled to plan meals around work schedules and available ingredients.
"If an app could tell me what healthy meal I can make quickly using foods I already have, I would use it daily."
Convenience + local availability is highly valuable.
Gym-goer, Port Harcourt, Nigeria
The user tracks protein intake manually and substitutes ingredients frequently due to pricing.
"I spend too much time searching for cheap protein alternatives."
Goal-based nutrition recommendations are valuable.
Parent, Ibadan, Nigeria
The user wanted balanced meals for children but lacked confidence in understanding nutrition.
"I just want to know if what my children are eating is actually balanced."
Nutrition explanation and confidence-building matter.
NYSC member, Enugu, Nigeria
The user often cooks based on whatever ingredients are cheapest nearby.
"I don't need fancy food. I need to know the smartest meal for my money."
Users optimise for value, not perfection.
- Tell me about the last time you had to decide what to eat with limited money or ingredients.
- What matters most when choosing food: price, taste, fullness, health, or speed?
- Have you ever wanted healthier meals but could not figure out what to cook?
- What makes nutrition apps unrealistic for you?
- If a system recommended meals based on your budget and nearby foods, would you trust it?
- What would make you stop using it?
flowchart TD
A[User] --> B[Next.js Web App]
B --> C[Authentication Layer]
B --> D[Meal Recommendation Engine]
D --> E[Recipe Database]
D --> F[Nutrition Knowledge Base]
D --> G[Local Ingredient & Price Layer]
D --> H[AI Reasoning Layer]
H --> I[Gemini API\nGoogle AI Studio Free Tier]
D --> J[Recommendation Results]
J --> K[Meal Plans]
J --> L[Nutrient Breakdown]
J --> M[Ingredient Alternatives]
J --> N[Budget Analysis]
B --> O[Analytics]
B --> P[Error Monitoring]
sequenceDiagram
participant U as User
participant W as Web App
participant A as Auth Service
participant R as Recommendation Engine
participant DB as Database
participant AI as Gemini API
U->>W: Enter budget, goals, ingredients, location
W->>A: Authenticate user
A-->>W: Session token
W->>R: Send user constraints
R->>DB: Fetch recipes and nutrition data
R->>AI: Generate optimized recommendations
AI-->>R: Ranked meal suggestions
R-->>W: Return recommendations
W-->>U: Show meals, nutrients, prices, alternatives
| Component | Technology |
|---|---|
| Frontend | Next.js |
| Hosting | Vercel |
| Backend | Supabase + Edge Functions |
| Database | PostgreSQL |
| Authentication | Supabase Auth |
| AI Layer | Gemini API (Google AI Studio) |
| Analytics | PostHog |
| Monitoring | Sentry |
| CDN / Security | Cloudflare |
| Repository | GitHub |
| CI/CD | GitHub Actions |
Authentication is handled using Supabase Auth.
Supported methods:
- Email/password
- Google sign-in (future)
Authentication is necessary for:
- Saving meal preferences
- Storing user goals
- Tracking recommendation history
- Personalising recommendations over time
- Budget
- Ingredients available
- Dietary preferences
- Nutrition goals
- User location
- Recipe matching
- Ingredient substitution
- Nutrition scoring
- Price optimisation
- AI-assisted ranking
- Meal recommendations
- Nutritional explanation
- Cost estimation
- Ingredient alternatives
- Meal preparation suggestions
We chose Next.js because:
- Fast development speed
- Excellent React ecosystem
- Easy deployment
- Server-side rendering support
- Strong developer experience
- Good performance for mobile-first applications
Vercel Hobby Plan
Why:
- Free deployments
- Preview deployments
- CI/CD integration
- Edge delivery
- Good support for student/startup projects
We chose Supabase because:
- PostgreSQL database included
- Built-in authentication
- Realtime support
- Edge functions
- Fast backend setup
- Easy integration with Next.js
Supabase Free Plan
Why:
- Free PostgreSQL database
- Free authentication
- Free storage
- Suitable for MVP-stage products
We chose PostgreSQL because:
- Structured relational data fits recipes and nutrition well
- Strong querying capabilities
- Reliable and scalable
- Excellent ecosystem support
The database stores:
- Recipes
- Ingredients
- Nutrition values
- User preferences
- Budget history
- Regional food mappings
We chose Gemini because:
- Google AI Studio provides a genuinely free tier with no credit card required
- Gemini 1.5 Flash is fast, cost-efficient, and capable for structured recommendation tasks
- Strong reasoning for contextual queries in the food and nutrition domain
- Natural language understanding of local food names and contexts
- Fully compliant with The Build's free-tiers-only rule
Google AI Studio Free Tier
Limits:
- 15 requests per minute
- 1,000,000 tokens per day
- 1,500 requests per day (Gemini 1.5 Flash)
The AI layer helps with:
- Meal optimisation
- Ingredient substitution
- Natural language food understanding
- Personalised recommendations
We minimise token costs by:
- Using retrieval before generation
- Caching common recommendations
- Keeping prompts compact
We chose Cloudflare because:
- Free CDN
- Free SSL
- Basic DDoS protection
- Fast global delivery
- DNS management
This helps keep the application fast and secure.
We chose PostHog because:
- Product analytics
- Session replay
- Funnel tracking
- User behaviour understanding
We will use analytics to:
- Understand recommendation quality
- Measure retention
- Track feature usage
- Improve meal suggestions
We chose Sentry because:
- Error tracking
- Performance monitoring
- Debugging support
- Production observability
This helps us identify failures quickly.
We chose GitHub Actions because:
- Free for public repositories
- Integrated into GitHub
- Easy automated testing
- Pull request validation
We will use it for:
- Running tests
- Linting
- Deployment workflows
AI-powered local meal intelligence for Africa.
Millions of Africans struggle daily to make healthy food decisions under real constraints such as:
- Budget
- Ingredient availability
- Limited nutrition knowledge
- Local food accessibility
Existing nutrition systems are often disconnected from African realities.
NutriLocal recommends balanced and affordable meals using:
- Local foods
- Local recipes
- User goals
- Budget constraints
- Ingredient availability
- Budget-aware meal recommendations
- Local ingredient understanding
- Nutrition scoring
- Ingredient substitution
- Goal-based suggestions
- Meal planning
- Next.js
- Supabase
- PostgreSQL
- Gemini API (Google AI Studio)
- Vercel
- PostHog
- Sentry
To become Africa's most intelligent local nutrition and meal recommendation platform.
- Problem definition
- User identification
- Validation conversations
- System architecture
- Tech stack planning
- Infrastructure planning
- Public repository initialisation
Week 2 goals:
- Launch first public MVP
- Build recommendation flow
- Deploy live application
- Configure CI/CD
- Add initial tests
- Implement authentication
- Create first recommendation dataset