Landing Page
Youtube video analysis with gemini 3 and yt v3 api
Youtube channel analysis with gemini 3 and yt v3 api
Agentic AI for Real Time insights of Influencer using gemini 3
Image Generation Using Gemini Model
Video Generation Using Veo Model

🚀 Kartr — AI-Powered Influencer–Sponsor Intelligence Platform

Turning chaotic influencer marketing into a structured, data-driven system.

Kartr is a full-stack AI platform that detects sponsorships, maps creator–brand relationships, automates campaigns, and generates social content using multimodal AI. Built for brands, agencies, and creators who want clarity instead of guesswork.

🌍 Inspiration

Influencer marketing is exploding, but the tools behind it still feel manual and fragmented. Brands struggle to track who is sponsoring whom, creators lack visibility into opportunities, and campaigns often run on guesswork instead of data.

Working closely with YouTube sponsorships exposed how inefficient and opaque this space is. Kartr was inspired by the idea of building an intelligence layer for the creator economy — one that can detect sponsorships, match brands with the right influencers, automate social media workflows, and generate content with AI.

By combining multimodal models, real-time search, and automation, the goal is to turn chaotic influencer marketing into a transparent, data-driven system.

⚡ What Kartr Does

Kartr connects brands and influencers through automated sponsorship intelligence and campaign execution.

🧠 Core Flow

Users log in with Google OAuth
Brands run bulk or single YouTube analysis
Gemini 3 detects sponsors and extracts brand relationships
Sponsors create campaigns
Agentic AI checks influencer credibility using Tavily + Gemini
Influencers accept campaigns
Generate content → images, videos, captions
Auto-post to social platforms
Track campaign analytics in real time

🤖 AI Capabilities

Sponsor detection from YouTube videos
Creator–brand relationship mapping
AI image generation → Gemini 2.5 Flash Image
AI video generation → Veo-2.0-generate-001
Caption + post automation
Real-time credibility checks
Campaign analytics + insights

Two Gemini 3 chatbots power the experience:

💬 General assistant for strategy and navigation
📊 Agentic assistant using Tavily for real-time data and campaign insights

All data is stored in Firebase and visualized in a unified dashboard.

🏗️ How We Built It

Kartr is deployed on Google Cloud with Dockerized services for scalability.

🔐 Authentication

Google OAuth login
Firebase Auth

📡 Data Collection

YouTube Data API → fetch channel & video data
Gemini 3 → analyze transcripts, metadata, and descriptions
Extract sponsors, brands, and niches

🧠 Intelligence Layer

Gemini-powered search for influencer matching
Tavily + Gemini → real-time credibility checks
Graph mapping of creator–brand relationships
Firebase stores structured data

🎬 Content Engine

Images → Gemini 2.5 Flash Image
Videos → Veo-2.0-generate-001
Caption generation
Auto-posting to Bluesky

💬 Chat Assistants

Chatbot 1 → General help
Chatbot 2 → Agentic real-time insights with Tavily

Everything runs through a React dashboard connected to a FastAPI backend.

🧩 Challenges We Ran Into

Detecting sponsorships reliably was harder than expected because creators disclose ads in many different ways — spoken mentions, description links, or subtle overlays. Building a system that understands all of these required combining transcript parsing, multimodal Gemini analysis, and custom logic.

Handling noisy YouTube data and API rate limits during bulk analysis also required optimization. Matching brands with the right influencers needed tuning to avoid irrelevant results.

Integrating multiple models for video, image, and reasoning created orchestration and latency challenges. Ensuring real-time Tavily insights while keeping the system fast meant carefully balancing cost, speed, and accuracy.

🏆 Accomplishments We're Proud Of

We built a working end-to-end platform that goes beyond analysis and actually connects brands and influencers into a usable workflow. Kartr can detect sponsors from real videos, map relationships, and match campaigns with relevant creators using Gemini-powered search.

We integrated multiple AI models for text, image, and video generation, added real-time credibility checks with Tavily, and deployed everything on Google Cloud using Docker.

The dual-chatbot system makes the platform feel like an intelligent assistant instead of just a dashboard. Turning a messy industry problem into a functional AI system within hackathon time is a major win.

📚 What We Learned

Multimodal AI is powerful but requires careful orchestration across models, APIs, and pipelines. Sponsorship detection involves context, audio, and patterns across videos — not just text.

We saw how fragmented influencer marketing data really is and how valuable structured insights can be. Working with Gemini, Tavily, and media generation models taught us how to balance speed, cost, and accuracy in real-time systems.

We also improved our skills in async backend design, cloud deployment, and AI integration. Building for real users forced us to constantly iterate between product needs, AI capability, and system architecture.

🛠️ Technologies Used

Frontend:
Bun.js, React, TypeScript, Tailwind CSS, shadcn/ui, Redux Toolkit, Framer Motion, Recharts, Axios

Backend & AI:
FastAPI (Python), Uvicorn, Firebase (DB + Auth), Gemini 3, Gemini 2.5 Flash (image generation), Veo-2.0-generate-001 (video generation), Tavily API (real-time data), YouTube Data API v3, NetworkX, Pandas , Bluesky API

Infrastructure & Orchestration:
Google Cloud Platform, Docker, async task orchestration, AI agent pipelines, RAG pipelines, REST APIs, cloud storage, background workers, scalable microservices architecture