Braindone.AI

Effortlessly create, manage, and publish AI-generated media content with our end-to-end automation platform powered by LLMs and Agentic AI.

Comment

📹 LLMs-to-Agentic-AI: Content Creation and Publishing Automation 🚀 Inspiration Our inspiration came from observing how many businesses, especially small and medium-sized ones, struggle to consistently create and publish engaging content on social media. With the rise of powerful Large Language Models (LLMs) and the emerging concept of Agentic AI, we saw a unique opportunity: fully automating the content pipeline from idea to publication. The "From LLMs to Agentic AI" hackathon at ESSEC, organized by Kryptosphere and Utopia, gave us the perfect environment to explore this vision.

🧠 What We Learned Throughout the project, we learned how to:

Integrate LLMs (like OpenAI GPT-4) to generate structured, high-quality content from simple prompts.

Apply Agentic AI principles to simulate decision-making for media selection and narration.

Handle automated multimedia creation using tools like FFmpeg.

Store and deliver assets efficiently using AWS S3 and CloudFront.

Connect to LinkedIn's API for seamless and automatic media publishing.

Build a modern and responsive frontend with React & TailAdmin.

🏗️ How We Built It 🔧 Stack Overview Frontend: React + TypeScript + TailAdmin

Backend: Node.js + Express

AI: OpenAI GPT-4 for script generation, Agentic logic for media composition

Media Processing: FFmpeg

Database: MySQL for metadata and user content

Cloud Infrastructure: AWS (S3, EC2, CloudFront)

Deployment: Docker + GitHub CI/CD

🧱 Step-by-Step Architecture User Input Users input prompts via a simple web interface.

Script Generation The backend uses OpenAI GPT to generate video scripts from prompts.

Automated Media Selection Agentic AI logic selects appropriate visuals, audio, and narration.

Media Creation & Processing Tools like FFmpeg help stitch together the components.

Cloud Storage & Management Final content is uploaded to AWS S3 and versioned with id_media.

Social Media Posting Content is automatically published via LinkedIn’s API with description and hashtags.

⚠️ Challenges We Faced Time Constraints: Building a full-stack project with automation in a short time was intense.

Image generation: Challenging because we didn't had enough power.

Cloud Infra: Challenging to design a complete architecture.

Agentic Logic: Creating decision flows that mimic human choices for media selection took iteration.

Cloud Permissions: Setting up proper AWS credentials and access policies was tricky during deployment.

🙌 Final Thoughts This project showed us how far we can go when combining creativity with cutting-edge technology. LLMs-to-Agentic-AI is just the beginning — we’re excited to expand it with more AI-driven customization, platform integrations, and analytics.

Built With

Share this project:

Updates