Brief - Autonomous Academic Agent

👉 OFFICIAL VIDEO DEMO LINK (CLICK HERE): [https://drive.google.com/file/d/1GGsmNc4_USdCbasvy1q5zqGHaYZxFECh/view?usp=drive_link]

Inspiration

As a future Information Technology student who just earned a full tuition-free scholarship, I wanted to solve a real academic problem. University lectures and courses are getting longer, and students are completely overwhelmed with information. I built "Brief" to act as an intelligent mentor, teaching students the most efficient, structured path to master the material by themselves.

What it does

Brief is an autonomous academic agent that operates beyond standard chat loops. A user uploads a PDF or PowerPoint lecture up to 200MB, and the system instantly delivers three core pillars: a high-level Executive Summary, an interactive Day-by-Day Action Plan for exams, and a simulated audio podcast script.

How I built it

The application features a clean Python and Streamlit frontend styled with a premium Cobalt theme. It utilizes the official Google GenAI SDK to connect directly with the Gemini 3.5 Flash API. To ensure global availability, the system is fully containerized using Docker and deployed on a secure, serverless Google Cloud Run infrastructure with strict resource scaling.

Challenges I ran into

Developing this agent in a limited timeframe required solving key architectural challenges. Deploying a Streamlit application with specific file-parsing libraries on Google Cloud Run meant building custom Docker layers and managing memory restrictions carefully. I also had to optimize the prompt flow to prevent timeout errors when processing heavy university slides while keeping the user interface completely responsive.

Accomplishments that I am proud of

I am incredibly proud of taking this project from a local prototype on my computer to a fully secure, scalable, live web production environment on Google Cloud Run. Successfully integrating the official Google GenAI SDK and a premium Cobalt-themed user interface, while generating real operational audio outputs, is a huge milestone for my second global hackathon.

What I learned

I learned how to structure autonomous software agents using the advanced capabilities of the Gemini 3.5 Flash API. I mastered containerization with Docker and learned how to scale cloud resources efficiently under real-world constraints. Most importantly, I understood how to analyze feedback and translate technical limitations into viable enterprise roadmaps.

What's next for Brief - Autonomous Academic Agent

While this version is a rapid hackathon MVP optimized for speed, it is natively prepared for enterprise scaling. In the next release, I will integrate an asynchronous Celery task queue to handle massive textbooks and a strict RAG (Retrieval-Augmented Generation) architecture to anchor all answers and eliminate any risk of academic hallucinations.

Built With

• cloudrun
• docker
• geminiapi
• google-cloud
• python
• streamlit