SimplyMail | Devpost

How we used GenAI for SimplyMail?

Inspiration

Email overload is a common problem, with inboxes flooded by long messages, spam, and potential phishing threats. We wanted to create a tool that simplifies email management while keeping users safe. SimplyMail was born out of a desire to make email communication more efficient and secure using AI-powered automation.

What it does

SimplyMail enhances the Gmail experience by summarizing emails directly within the Gmail tab, allowing users to grasp key information at a glance. It also scans for harmful or phishing links and alerts users to potential security threats, helping them stay protected from scams and cyberattacks.

How we built it

We combined several technologies to bring SimplyMail to life:

Flask for the backend, handling API requests and communication between components.
MongoDB as our database, storing summaries and flagged security threats. To optimize performance, we query the database to check if a summary has already been generated for an email before generating a new one, reducing unnecessary processing.
Ollama and DeepSeek (LLM) for generating accurate and concise email summaries locally, ensuring privacy.
nGrok to expose our local server for testing and development.

Challenges we ran into

Optimizing AI Summaries: Ensuring the locally run LLM produced useful and concise summaries without losing critical details required fine-tuning.
Phishing Detection Accuracy: Identifying malicious links without false positives was a challenge that required refining our detection logic.
Frontend-Backend Connection Issues: Connecting our JavaScript-based extension to the Flask backend was more difficult than expected, especially due to the restrictive nature of the provided WiFi, which blocked certain requests and connections. Troubleshooting this issue involved experimenting with nGrok, CORS configurations, and alternative network solutions.
Efficient Summary Storage & Retrieval: We wanted to avoid regenerating summaries for emails that had already been processed. Using MongoDB queries, we implemented a caching mechanism that checks for existing summaries before generating new ones, optimizing for time and resources.

Accomplishments that we're proud of

Successfully integrating AI-driven summarization directly within Gmail.
Building a functional phishing detection system that alerts users to potential risks.
Overcoming network restrictions to establish communication between the frontend and backend.
Implementing efficient database querying to optimize response times and reduce redundant processing.
Learning and implementing new tools like MongoDB, Flask, and Ollama in a short timeframe.
Creating a seamless user experience with real-time email analysis.

What we learned

How to work with Flask to build a backend API.
The fundamentals of MongoDB for managing and storing data efficiently.
Setting up and running LLMs locally to enhance privacy and performance.
Using nGrok to facilitate testing and development.
Troubleshooting network-related issues when connecting a frontend extension to a backend service.
The importance of caching and database optimization in performance-critical applications.
The role of user experience in security-focused applications.

What's next for SimplyMail

We plan to enhance SimplyMail with:

Better AI Summaries: Fine-tuning DeepSeek for improved context understanding.
Real-Time Phishing Alerts: Expanding detection capabilities for more advanced threats.
Browser Extension Support: Making SimplyMail available as a Chrome extension on the online extension store for easier access.
Multi-Platform Compatibility: Extending support beyond Gmail to other email providers.

SimplyMail is just the beginning. Our goal is to make email smarter, safer, and more efficient for everyone! 🚀