The latest articles on DEV Community by Kenneth Ansel (@kennethansel). https://dev.to/kennethansel https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3721173%2F19f63301-1c3d-4a57-803e-53b9ad560a21.jpeg https://dev.to/kennethansel en Kenneth Ansel Fri, 28 Aug 2026 10:08:42 +0000 https://dev.to/kennethansel/building-collabkar-an-ai-powered-influencer-marketing-platform-with-gemini-950 https://dev.to/kennethansel/building-collabkar-an-ai-powered-influencer-marketing-platform-with-gemini-950 <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F28ydr36mf7nzllm5dg36.jpg" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F28ydr36mf7nzllm5dg36.jpg" alt=" " width="799" height="351"></a><em>This post was created for the purposes of entering the All Things Agentic Hackathon (Google Cloud &amp; Gemini).</em></p> <h2> The problem </h2> <p>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 <em>fair</em> rate for a campaign? And once you pay, is your money actually safe?</p> <p>That's the gap <strong>CollabKar</strong> sets out to close.</p> <h2> What CollabKar does </h2> <p>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.</p> <p>Three things make this possible:</p> <ul> <li> <strong>Natural-language creator matching</strong> — powered by <code>gemini-3.5-flash</code>, so brands don't need to know the right search filters, they just describe what they want</li> <li> <strong>Fair-price rate benchmarking</strong> in INR, so neither side is negotiating blind</li> <li> <strong>Fake-follower detection</strong> to flag inflated audiences before a brand commits budget</li> </ul> <h2> How it's built </h2> <ul> <li> <strong>Frontend:</strong> Next.js (App Router) + Tailwind CSS + Framer Motion, deployed on Vercel</li> <li> <strong>Backend:</strong> Node.js + Express, containerized with Docker, deployed on Render</li> <li> <strong>Data layer:</strong> MongoDB (via Mongoose) + Redis for caching</li> <li> <strong>AI layer:</strong> Gemini API (<code>gemini-3.5-flash</code>) for natural-language matching and pricing suggestions</li> </ul> <h2> What was tricky </h2> <p>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.</p> <h2> Try it </h2> <ul> <li>🔗 Live demo: <a href="https://collab-l6o7.vercel.app/" rel="noopener noreferrer">https://collab-l6o7.vercel.app/</a> </li> <li>💻 Code: <a href="https://github.com/KennethAnsel/Collabkar_New" rel="noopener noreferrer">https://github.com/KennethAnsel/Collabkar_New</a> </li> </ul> <p>Built for the <strong>All Things Agentic Hackathon</strong> by Google Cloud &amp; Gemini.</p> ai webdev gemini hackathon