PROJECT_01
NexTask — Kanban Board
AI Powered · Drag & Drop · Real-time UI
A clean, responsive AI powered Kanban board app with drag-and-drop task management, column-based workflow, and real-time UI updates. Built with a focus on smooth UX and modern frontend patterns.
React
JavaScript
CSS3
Vercel
PROJECT_02
Pulsar Star Detection as a Service
99% Accuracy · ROC-AUC 0.973
Cloud-deployed ML microservice using FastAPI + LightGBM to classify pulsar candidates from telescope features. Streamlit dashboard with batch predictions, EDA visualizations, and SHAP-based interpretability.
FastAPI
LightGBM
Streamlit
Docker
AWS EC2
PostgreSQL
Optuna
PROJECT_03
Cross-lingual Sentiment Analysis
Zero-shot & Fine-tuned · 3 Transformer Models
Multilingual emotion-classification workflows using mBERT, XLM-R, and XLNet. Evaluated across multi-domain datasets with reproducible benchmarking pipelines and optimized tokenization strategies.
PyTorch
HuggingFace
mBERT
XLM-R
XLNet
Scikit-learn
PROJECT_04
Chest X-Ray Classification CNN
AUC 0.99 COVID · 0.98 Pneumonia · 0.97 Normal
CNN models to classify chest X-rays into COVID-19, Pneumonia, and Normal categories. Applied data augmentation, EDA, and GPU-accelerated training to reduce overfitting on imbalanced medical imaging datasets.
TensorFlow
Keras
OpenCV
CNN