Full Stack Development | Generative AI | Building Scalable AI-Powered Applications

Sapna Tiwari  • Peerlist

Built a full-stack habit-tracking Progressive Web App to solve a problem most habit trackers ignore — people don't quit because the app is bad, they quit because one missed day feels like failure, and they never come back. AI Habit Tracker is designed around that moment, using Google Gemini 2.5 Flash to power five personalized AI features: streak-recovery coaching, an interactive habit analysis chat, weekly progress reports, personalized habit suggestions, and daily morning motivation — each generated using the user's real habit and streak data, not generic templates. On the backend, I built secure REST APIs with Node.js and Express, with JWT and bcrypt handling authentication, and MongoDB Atlas (via Mongoose) managing cloud persistence — with date-fns handling the streak-calculation logic to keep date math reliable across time zones and edge cases. To keep users engaged without being intrusive, I implemented PWA capabilities with Web Push notifications secured via VAPID authentication, paired with a cron job that monitors daily habit completion and triggers automated streak reminders before a streak actually breaks. This project pushed me to think beyond "calling an AI API" — the real engineering challenge was deciding what context to feed into each prompt so the responses felt specific and useful rather than generic, while also handling real backend concerns like authentication, scheduled jobs, and cloud data persistence.

AI Habit Tracker

A real-time collaborative code editor to solve a problem I kept running into myself — pair programming with friends remotely always meant fighting with screen-share lag or passing code back and forth over chat. DevCollab fixes that by letting multiple people edit the same file live, using Monaco Editor (the same editor that powers VS Code) paired with Socket.io for real-time synchronization, complete with live cursor tracking and role-based access control (owner/editor/viewer) so teams can control who can edit versus just view.Beyond live editing, I added peer-to-peer video and voice calling using WebRTC, so collaborators can talk while they code without switching to a separate app. On the backend, I implemented JWT-based authentication with bcrypt password hashing and refresh tokens for persistent, secure sessions, and designed a MongoDB schema to support versioned code history — so users can track changes and roll back to an earlier version if needed. I also integrated Razorpay for payments, with server-side signature verification to securely unlock premium features.The most interesting engineering challenge here wasn't broadcasting changes with Socket.io — it was thinking through session management, authorization logic for different user roles, and how to keep the editor feeling responsive while syncing state across multiple users in real time.

DevCollab — Real-Time Collaborative Code Edit

Built a full-stack social posting application using the MERN stack where users can create, edit, and delete their own posts, with secure JWT-based authentication ensuring users can only modify content they own. Beyond basic CRUD, I implemented a comments system so users can engage in discussions on posts, and a like feature to let users react to content they enjoy — the two interactions that turn a static feed into something people actually want to come back to. On the backend, I designed RESTful APIs with Node.js and Express, with MongoDB and Mongoose handling data modeling for posts, comments, and likes, including the relationships between them (e.g., tracking which users liked or commented on which posts). Authentication and authorization were handled with JWT, ensuring routes like editing or deleting a post are protected and scoped to the post's actual owner — not just gated behind a login check. Working on PostFlow helped me get comfortable with the fundamentals that show up in almost every real-world application: structuring a relational-style schema in a NoSQL database, designing protected API routes, and building a clean CRUD flow end-to-end — skills that became the foundation for the more complex features I added in later projects like DevCollab and AI Habit Tracker.