Projects

Deep Learning

Deep learning–powered web app for classifying 13 types of Bangladeshi flowers.
Transfer learning with MobileNetV2 (fine-tuned last 50 layers) on a custom curated dataset.
Advanced data augmentation pipeline (rotation, shift, shear, zoom, flip) to improve model robustness.
Achieved 99.87% validation accuracy with real-time classification and confidence scores via a Streamlit app.

Tools / Technology: Python, TensorFlow/Keras, Streamlit, OpenCV, PIL, NumPy, Pandas

NLP

Analyze social media posts to classify sentiment (positive, negative, neutral) using transformer models.
Implemented custom tokenization and feature-extraction pipelines for multilingual text processing.
Deployed an interactive dashboard for real-time sentiment analysis with visualization of trends.

Tools / Technology: Python, Hugging Face Transformers, spaCy, NLTK, BERT, RoBERTa

Machine Learning

Detect anomalies in network traffic for cybersecurity, performance monitoring, and compliance.
Isolation Forest (unsupervised ML) with temporal feature engineering for pattern recognition.
Developed an automated alert system with configurable thresholds and notification mechanisms.

Tools / Technology: Python, Scikit-learn, Pandas, Streamlit, Plotly

Machine Learning

Comprehensive movie recommendation system with content-based and collaborative filtering (KNN, SVD).
Implemented a hybrid approach combining multiple algorithms with weighted averaging for robust recommendations.
Built an interactive Streamlit web application deployed on the cloud for real-time movie suggestions.
Achieved RMSE of 0.454 with the SVD model, a 58% improvement over the baseline KNN approach.

Tools / Technology: Python, Streamlit, Scikit-learn, Pandas, NumPy, TruncatedSVD

Data Science

Analyzed global energy consumption patterns using advanced regression models to predict future energy usage.
Created interactive visualizations to illustrate energy usage trends across different regions and time periods.
Implemented a Streamlit web application for user-friendly data exploration and prediction.

Tools / Technology: Python, Scikit-learn, Pandas, Plotly, Streamlit

Machine Learning

Forecast store sales using regression and time-series techniques on historical sales data.
Interactive dashboard comparing forecasted versus actual sales for performance monitoring.

Tools / Technology: Python, Scikit-learn, Pandas, Plotly, Streamlit

Machine Learning

Predict whether a telecom customer is likely to churn using logistic regression and ML classifiers.
Deployed as a Streamlit web app for interactive, real-time churn prediction.

Tools / Technology: Python, Scikit-learn, Pandas, Streamlit