Jamala

Inspiration

In Yoruba we say: “Je Amala, kí ọ̀nà rẹ lè lá” — Eat Amala so that your path may open.
Amala is more than food; it’s culture, history, and community. Yet finding authentic Amala spots is still mostly word-of-mouth. We built JAMALA to preserve food heritage and make discovery easier.

What it does

JAMALA is a crowdsourced map where users can submit, verify, and discover Amala spots.

• Map UI with clustering, filters, and reviews.
  
• Agentic intake (chat/voice) to add spots quickly.
  
• Autonomous discovery scans the public web for likely Amala locations, then routes them into a verification queue.
  
• Open source so it lives on beyond the hackathon.

How we built it

• Frontend: Next.js + Tailwind on Vercel.
  
• Maps: MapLibre for geosearch, clustering, and performance.
  
• Backend: Firebase (Auth, Firestore, Storage).
  
• Agents: Google Agent Development Kit + GPT for chat/voice intake.
  
• Web crawlers for auto-discovery and candidate scoring.
  
• GitHub for version control and collaboration.

Challenges we ran into

• Shipping a usable MVP under 12 hours.
  
• Preventing duplicate entries from both users and crawlers.
  
• Making chat/voice agents responsive on weak networks.
  
• Validating crawler data so only true Amala spots get listed.
  

Accomplishments that we're proud of

• Delivered a working MVP with live demo link.
  
• Built a verification flow combining autonomy and community trust.
  
• Integrated A2A agent orchestration for food mapping.
  
• Turned a Yoruba proverb into a brand identity that resonates.
  

What we learned

• How cultural context strengthens product adoption.
  
• The importance of a Definition of Done for hackathon speed.
  
• Technical lessons in maps, AI orchestration, and data validation.
  
• That a prototype in users’ hands beats a slide deck every time.
  

What's next for Jamala

• Expand autonomous discovery across blogs, directories, and social media.
  
• Add gamified verification to reward community validators.
  
• Launch a mobile-first app for quick discovery on the go.
  
• Extend the model to other cultural foods (e.g., Fufu, Jollof, Efo Riro).
  
• Secure a domain + OSS repo so JAMALA grows beyond this hackathon.
  

Built With

• cheerio/playwright
• deepgram
• elevenlabs
• firebase-(auth/firestore/storage/cf)
• geojson/supercluster
• github
• google-adk-+-a2a
• maplibre
• next.js
• openai-gpt
• shadcn/ui
• tailwind
• typescript
• vercel
• vercel-functions