A collection of Jupyter notebooks covering the fundamentals of Digital Signal Processing (DSP) for audio analysis and machine learning using Python and Librosa.

Vision Transformer based image classification project using the Dogs vs Cats Kaggle dataset with PyTorch.

Basic CNN implementation for CIFAR-10 image classification using TensorFlow and Keras.

A modular RAG experimentation framework benchmarking BM25, vector, hybrid, and reranked retrieval using Recall@1, Recall@5, and MRR, designed as a foundation for future agentic RAG systems.

Fine-tunes LLaMA-2 using QLoRA for instruction-style text generation, demonstrating dataset formatting, parameter-efficient LLM training, and iterative inference debugging to achieve clean and stab…

Implements LoRA and QLoRA parameter-efficient fine-tuning pipelines for LLaMA-2, demonstrating memory-efficient LLM adaptation workflows using HuggingFace Transformers and PEFT.

Fine-tunes the GPT-2 language model on Shakespearean text to generate stylistically consistent dialogue, demonstrating domain adaptation, loss and perplexity tracking, and transformer-based text ge…

SMS spam classification using a Transformer-based model built with HuggingFace and PyTorch, demonstrating modern NLP techniques for contextual text classification and model evaluation using a confu…

Compares LSTM and Bidirectional LSTM models for fake news detection, highlighting how bidirectional context improves text classification performance using deep learning.

Detects anomalies using the DBSCAN algorithm by identifying low-density noise points, with clear visualizations of original data, highlighted outliers, and isolated anomaly points.

Builds a spam detection model using machine learning to classify messages as spam or not spam, demonstrating text preprocessing, feature extraction, and model evaluation.

Detects anomalies using the Isolation Forest algorithm, with clear visual comparison between original data and anomaly-marked data in an unsupervised learning setup.

Detects anomalies using the Local Outlier Factor (LOF) algorithm, with clear visualizations of normal data, highlighted outliers, and isolated anomaly points.

Applies DBSCAN clustering on two different datasets to identify density-based clusters and compare clustering quality using silhouette scores and visual analysis.

Implements hierarchical clustering to group data points based on similarity, with clear visualization of cluster formation using an unsupervised learning approach.

Implements K-Means clustering using Silhouette Score and Elbow Method to identify the optimal number of clusters, with clear visualizations of clustering results and evaluation curves.

Predicts dining time using machine learning and feature engineering, achieving high classification accuracy with a detailed performance evaluation.

This project demonstrates Decision Tree Classification using two pruning techniques: pre-pruning and post-pruning. Both approaches are implemented and compared to control overfitting and improve mo…

This project uses a Decision Tree Regression model to predict housing prices from structured data. The model is trained and evaluated using RMSE and R² score, with visual analysis of residuals and …

A multiclass classification project using Logistic Regression with Python and scikit-learn. The model is evaluated using a classification report and achieves perfect precision, recall, and F1-score…

Implemented ElasticNet Regression, combining L1 and L2 regularization to handle multicollinearity and improve robustness. Evaluated using R² score (~0.64) and residual distribution analysis on real…

Developed a Lasso Regression model applying L1 regularization for housing price prediction. The model achieved an R² score of ~0.64 while performing implicit feature selection and maintaining stron…

Built a Ridge Regression model for housing price prediction to reduce overfitting through L2 regularization. Achieved an R² score of ~0.64, with residual analysis confirming improved model stability.

Implemented Linear Regression using scikit-learn to predict housing prices. Evaluated model performance using R² score (~0.64) and residual distribution analysis to assess prediction errors and reg…