This post was created for the purposes of entering the All Things Agentic Hackathon (Google Cloud & Gemini).
The problem
Influencer marketing in India is booming, but for MSMEs and D2C founders it's still a black box. How do you know if a creator's followers are real? What's a fair rate for a campaign? And once you pay, is your money actually safe?
That's the gap CollabKar sets out to close.
What CollabKar does
CollabKar is an AI-powered influencer marketplace that connects brands with vetted creators — no spreadsheets, no cold DMs, no guesswork. A brand can describe their campaign in plain English and get matched with relevant creators instantly.
Three things make this possible:
-
Natural-language creator matching — powered by
gemini-3.5-flash, so brands don't need to know the right search filters, they just describe what they want - Fair-price rate benchmarking in INR, so neither side is negotiating blind
- Fake-follower detection to flag inflated audiences before a brand commits budget
How it's built
- Frontend: Next.js (App Router) + Tailwind CSS + Framer Motion, deployed on Vercel
- Backend: Node.js + Express, containerized with Docker, deployed on Render
- Data layer: MongoDB (via Mongoose) + Redis for caching
-
AI layer: Gemini API (
gemini-3.5-flash) for natural-language matching and pricing suggestions
What was tricky
Getting the AI matching to feel conversational — "just describe your campaign" — while still surfacing enough concrete data (rates, engagement, niche fit) for a brand to actually trust the recommendation took real iteration. Fraud detection was its own challenge too: a sudden spike in a creator's followers isn't always fake, sometimes it's just a creator having a great week, and the system has to tell the difference.
Try it
- 🔗 Live demo: https://collab-l6o7.vercel.app/
- 💻 Code: https://github.com/KennethAnsel/Collabkar_New
Built for the All Things Agentic Hackathon by Google Cloud & Gemini.
Top comments (0)