π§ Inspiration We saw a rising problem β talented professionals and founders wanted to grow on LinkedIn, but either lacked consistency, content ideas, or time to post. While AI tools existed, none offered a truly end-to-end system that handled scraping, writing, humanizing, posting, replying, and analytics β all autonomously.
Thatβs where DIVIT.AI comes in β your personal AI Social Media Manager β‘
π€ What it does DIVIT.AI is a fully automated LinkedIn growth agent that: -Scrapes and clones your LinkedIn persona and writing style -Generates AI-to-Human undetectable posts -Autopublishes posts on your behalf -Auto-replies to comments with context-aware responses -Tracks engagement history and analytics -Supports real/mocked login, role switching (agency/client), and credit-based usage system -All with a smooth, interactive, 3D-infused UI inspired by Greek gods π
π οΈ How we built it Frontend: Built in Bolt using shadcn/ui + TailwindCSS + TypeScript Backend: Supabase (PostgreSQL) for auth, storage, and data ops AI Layer: Gemini + Groq backend for blazing fast human-like generations Integrations: LinkedIn scraping + API flow LYZR agent models for persona-based writing OpenRouter fallback Credit simulation and analytics visualized within dashboard
Design: Dark futuristic aesthetic with glassmorphism, scroll-based animations, 3D Spline scenes, and sparkles
π§ββοΈ Challenges we ran into Making the posts undetectable as AI (GPTZero, etc.) while still retaining human tone Handling LinkedIn scraping limits and rotation Integrating dynamic agent testing via form while auto-submitting metadata Creating real-time, role-based UI states in Bolt (buggy at first) Keeping design luxurious yet performant (lazy loading + 3D)
π Accomplishments that we're proud of Generated 100+ human-like posts with 90%+ GPT detection pass rate Built a working Universal Creator Form that lets anyone upload their own AI agents Built mocked and real AI workflows for the platform Seamlessly integrated Gemini as chatbot that queries entire platform data Auto-replying comments agent tested and working inside Cursor
π What we learned How to work with Boltβs limitations and strengths How to simulate real AI pipelines inside a no-code setup How to convert design-heavy interfaces into functional growth engines Importance of testing AI output against detection and personalisation
π What's next for DIVIT.AI Move from mocked β live deployment with real API keys & connections Launch a creator marketplace for AI agents (rent, sell, plug-n-play) Introduce a LinkedIn CRM to manage leads directly inside the dashboard Expand to Twitter & Instagram automation Enable team-level access and collaboration for agency accounts
DIVIT.AI isn't just a tool. It's an ecosystem for creators, founders, and teams who want to win on LinkedIn β without ever lifting a finger.
Built With
- bolt
- framer-motion
- gemini-api
- gptzero
- groq
- linkedin-unofficial-api
- lyzr
- openrouter
- postgresql
- react
- shadcn/ui
- splinetool/react-spline
- supabase
- tailwindcss
- testing
- tsparticles/react
- typescript
- vite
Log in or sign up for Devpost to join the conversation.