Project Vault

All my projects collected in one place.

Histopathology Image Classification

Built and tuned SVM, MLP, and CNN models for colorectal cancer tissue classification.
Achieved validation accuracy of 72% with stratified 5-fold CV and model tuning using Keras Tuner.
Conducted error analysis and ensured experiment reproducibility.
PythonTensorFlowscikit-learnMLPCNNKeras Tuner

Image Captioning Web App – AWS Full Stack Deployment

Deployed a scalable Flask web app using EC2 Auto Scaling and ALB for high availability.
Orchestrated a serverless backend with Lambda and EventBridge for image captioning using the Gemini API and thumbnail generation.
Built an S3-triggered pipeline to generate captions and thumbnails, storing results in RDS and returning pre-signed URLs.
Configured IAM, VPC, and CloudWatch for secure, observable infrastructure.
FlaskAWS EC2S3RDSLambdaEventBridgeIAMAuto ScalingALBPython

Pulmonary Disease Classification

Developed a Deep Learning Framework to classify pulmonary diseases through respiratory sound analysis.
Transformed raw audio files into spectrographic images and employed CNN for accurate detection across 8 classes.
PythonDeep LearningCNNEfficientNetB0

Mood-Based Music Recommender System

Built a real-time playlist recommender integrating facial emotion detection and sentiment analysis.
Mapped emotional states to Spotify’s track features (valence, tempo) to generate dynamic playlists.
Deployed an interactive UI using Streamlit for user engagement.
DenseNet121StreamlitHugging FaceDistilBERTSpotify API

Efficient ETL Pipeline

Developed a streamlined ETL (Extract, Transform, Load) pipeline to enhance data transfer and visualization via a web dashboard, improving error reporting using AWS services.
AWS RedshiftAWS GlueCloudWatchQuickSight

Customer Segmentation

Employed clustering algorithms to analyze customer behavior and spending patterns, resulting in a segmentation model that enables targeted marketing strategies.
ExcelPythonPandasNumPyMatplotlibSeaborn

Detection of Mental Health

Predicted mental health for adults using machine learning models like Logistic Regression, Decision Trees, Random Forest, KNN, and SVM on organizational employee data.
Achieved 84.
23% accuracy with cross-validation and feature selection.
PythonFlaskMachine LearningSklearnSeabornPandasPickle

Smart Lender

Designed and implemented a predictive web application assessing loan eligibility by comparing applicant details with predefined criteria.
Integrated analytics dashboards and chatbot support using IBM Watson.
PythonFlaskMachine LearningDecision TreeRandom ForestKNNXGBoostIBM CloudIBM WatsonJira

Crop Mapping Using Fused Optical-Radar Data Set

Developed and optimized machine learning models, including LDA, Logistic Regression, KNN, SVM, and Random Forest, for multi-class classification of 7 crop types.
Improved accuracy through effective preprocessing and hyperparameter tuning.
RMachine Learning

Project Management System - Bhumi

Developed a WhatsApp Proof of Concept for Tech4Change using Twilio and Streamlit, enabling users to submit service requests seamlessly.
Implemented automated messaging for efficient communication.
PythonTwilioStreamlit

Age and Gender Detection

Built a model to predict age and gender from images or webcam streams using computer vision techniques and Convolutional Neural Networks (CNN) with CaffeNet architecture.
Computer VisionConvolutional Neural NetworksCaffeNet