DrugLytics

🏆 Inspiration

Every day, millions of people mix medications without knowing if it's safe. Online tools are overwhelming, text-heavy, and impersonal. A doctor’s advice would be ideal, but long wait times and rising healthcare costs make quick consultations increasingly difficult.

We wanted to create a faster, more personalized, and interactive solution—one that simplifies drug safety, explains risks visually, and helps users make informed decisions instantly.

🚀 What It Does

User Profile Creation – Users enter the medications they take and any allergies they have.
Drug Interaction Search – Users search for a new drug to check for possible interactions.
Data-Driven Insights & Analytics – The system leverages a database of 300K+ drug interactions, analyzing trends, common searches, and high-risk drug combinations to improve recommendations and response accuracy.
Instant AI-Powered Insights – The system provides:
- A clear, simplified text summary of any interactions.
- A generated video animation showing the *interaction details and narrated warnings * in an easily digestible manner.
AI Chatbot for Further Discussion – Users can ask questions about their medications and get real-time responses.

Our goal? Make medication safety easy to understand, personalized, and accessible in seconds.

🛠️ How We Built It

We combined Data Analysis, GenAI, cloud storage, and molecular visualization to make DrugLytics powerful and scalable:

Dataset – We used the DrugBank dataset (300K+ drug interactions).
AI Multi-Agent System – We built an AI multi-agent using Google Gemini to search for additional drug-related information and generate video content.
Chatbot & Summaries – OpenAI API (ChatGPT) powers both the chatbot and the simplified text summaries.
Frontend & Backend – Built with Streamlit (frontend) and Flask (backend), hosted on AWS EC2.
Database & Caching – We used MongoDB Atlas to store interaction search data and index frequently searched drugs for faster results.
Molecular Visualization – We used RDKit to convert SMILES strings (chemical structures) into 3D molecule renderings and then used Manim to animate the results.
Cloud Storage for AI Videos – Cloudflare R2 stores previously generated videos so users don’t have to wait for repeated searches.
Secure Authentication – We integrated Okta Auth0 for secure, user-specific medication tracking.

🔥 Challenges We Overcame

1️⃣ Handling a Massive 300K+ Interaction Dataset Efficiently

Problem: Searching through such a large dataset slowed down performance.
Solution: We indexed common searches with MongoDB Atlas, making responses lightning-fast.

2️⃣ Converting Drug Data into Visual, Engaging Content

Problem: The dataset used SMILES strings, a text format for molecules that’s hard to visualize.
Solution: We used RDKit to generate 3D molecule structures and Manim to animate** them for video explanations.

3️⃣ Speeding Up AI Video & Text Generation

Problem: Generating a video and summary for every new interaction took too long.
Solution: Cloudflare R2 caching + MongoDB Atlas indexing reduced wait times for frequently searched drugs.

4️⃣ Ensuring Security for Users’ Medication Data

Problem: Users need to store personal medical information securely.
Solution: We integrated Okta Auth0 for secure authentication and encrypted storage.

🎯 Accomplishments We’re Proud Of

✅ Successfully integrated multiple AI models to create an intelligent, automated medical assistant.
✅ Indexed a massive medical dataset for real-time search capabilities.
✅ First-time use of Cloudflare R2, AWS EC2, and Okta Auth0—successfully implementing cloud-first solutions.
✅ Developed a molecular animation system that converts raw chemical data into user-friendly, AI-generated videos.

📚 What We Learned

How to optimize large-scale datasets for fast AI-driven querying.
How to integrate multi-agent AI systems for enhanced search and video generation.
How caching AI-generated media reduces infrastructure costs and improves performance.
The power of combining AI, cloud storage, and GenAI to create a seamless, interactive user experience.

🔮 What’s Next for DrugLytics?

🔹 More Personalization – Users will be able to input age, gender, weight, and medical history for more precise recommendations.
🔹 Medical Report Uploads – Users will be able to upload PDF medical records, and AI will extract relevant drug interactions.
🔹 Enhanced Video Summaries – We'll refine AI-generated explanations to make them even simpler and easier to understand.

🏆 Project Achievements

✅ We're using AI to improve healthcare education, helping patients make more educated health decisions.
✅ We were innovative in our use of generative AI and data analysis, using AI agents to turn complex workflows with chemical and video data into easy insights for users.
✅ We powered our frontend experience using Streamlit, making it smooth, fast, and user-friendly.
✅ We indexed high-traffic queries with MongoDB Atlas, prioritizing efficiency and yielding rapid search results from a large dataset.
✅ Users’ medication data is stored securely with authentication and encryption using Okta Auth0.
✅ Our Flask backend is deployed on AWS EC2, ensuring high availability and performance.
✅ By storing videos in Cloudflare R2, we eliminated long AI generation times for repeat searches.