Leaf Care — Plant Disease Detection
Flutter, Dart, Python, FastAPI, TensorFlow Lite, EfficientNet, Clean Architecture, Firebase
Bachelor of Computer Application final year project, Saraswoti Multiple Campus- Built Leaf Care as the final year project of my BCA: a Flutter app that diagnoses plant disease from a leaf photo. A FastAPI microservice runs an EfficientNet convolutional network exported to TensorFlow Lite (~70 MB) across 38 crop-disease classes and returns the prediction as JSON.
- Owned the full path from image to answer: capture or pick a leaf on device, resize and normalize it to the model input shape, run inference through the TFLite interpreter, then read the argmax class and confidence back in the app.
- Added confidence-aware treatment guidance resolved from a bundled disease and treatment database: at or above 50% confidence the app serves mild care steps, below it switches to an emergency workflow.
- Structured the Flutter client under Clean Architecture with separate presentation and data layers, Firebase Authentication, an agriculture and medicine reference library, a plant diary, and a seed and fertilizer shop.
SSTFlix — Movie Recommendation System
Flutter, Dart, Python, Flask, scikit-learn, TF-IDF, Cosine Similarity, Naive Bayes, Firebase, Riverpod, GetX, TMDB API
Bachelor of Computer Application, 6th semester (third year) project, Saraswoti Multiple Campus- Built SSTFlix as my third year (6th semester) project: a full-stack movie discovery app with a Flutter client for Android and iOS over a Python and Flask machine-learning backend deployed on Render behind Gunicorn.
- Combined two recommender strategies behind one endpoint, content-based recommendations from TF-IDF vectors of title and genre scored by cosine similarity, and collaborative filtering over a user-item rating matrix using cosine distance.
- Trained a TF-IDF plus Multinomial Naive Bayes classifier that labels TMDB reviews Good or Bad, and exposed cosine, Euclidean, and Manhattan similarity scores between any two movies.
- Integrated the TMDB API for details, posters, trailers, and a swipe-to-discover deck, with Firebase Auth and Firestore for accounts, Riverpod and GetX for state and routing, and SharedPreferences for the watchlist.
Fruit E-Commerce Seller Platform
PHP, MySQL, PDO, HTML, CSS, JavaScript
Bachelor of Computer Application, 4th semester project (with a partner), Saraswoti Multiple Campus- A short college CRUD project built with a partner in my 4th semester: a multi-vendor fruit marketplace in core PHP and MySQL.
- Sellers register and manage their own fruit listings with add, view, edit, delete, and a publish or draft status, while an admin panel moderates sellers and products.