Mailmind-Ai

Inspiration

Every day we drown in long, cluttered emails — important deadlines buried under fluff, and phishing attempts that look just convincing enough to trust. We wanted to build something that reads an email the way a smart assistant would: instantly understand it, flag what matters, and help you respond without wasting time.

What it does

MailMind AI takes any email and instantly delivers a summary, priority level (Low/Medium/High), category, sentiment, extracted tasks and deadlines, and a phishing/security risk score. It also generates ready-to-send smart replies in Professional, Friendly, and Concise tones — or a custom reply based on the user's preferred tone.

How we built it

The frontend is built with React, TypeScript, and Vite for a fast, responsive UI. The backend uses Node.js/Express, deployed as serverless functions on Vercel. All the intelligence comes from the Gemini API, guided by carefully engineered prompts that return structured, constrained JSON output for consistent, reliable results.

Challenges we ran into

Deploying the backend correctly on Vercel was tricky — we had to restructure our Express server into a serverless-compatible API route. We also hit a deprecated Gemini model mid-deployment and had to quickly switch models without breaking the JSON response structure our frontend depended on.

Accomplishments that we're proud of

Shipping a fully working, live-deployed product under intense time pressure. We're especially proud of the 95/100 security score and 98/100 problem statement alignment in our AI evaluation, and finishing 8th out of 198 participants.

What we learned

We learned how much structured prompting and strict output constraints improve AI reliability, and how critical proper environment configuration and serverless architecture are for a smooth deployment. We also learned that real debugging skills show up fastest under a ticking clock.

What's next for Mailmind-Ai

Next, we want to add direct Gmail/Outlook inbox integration, batch email processing, a browser extension for in-inbox suggestions, and multi-language support — turning MailMind AI from a standalone tool into a true email co-pilot.

MailMind AI was originally built during a college-level hackathon (Prompt Wars) at Goel Institute of Technology and Management, Lucknow, where it ranked 8th out of 198 participants. We're submitting it here to showcase the project to a wider audience and continue building on it.

Built With

• express.js
• gemini
• javascript
• node.js
• prompt
• react
• tailwind
• typescript
• vercel
• vite