AI-Learning-Journey

A curated portfolio showcasing my journey in artificial intelligence and machine learning. This repository contains a collection of projects demonstrating my skills in areas like natural language processing, computer vision, and data analysis.

Recent Activities (as of 2025-09-15)

• Added SQL Student Mental Health Analysis, demonstrating comprehensive data analysis using PostgreSQL to examine international students' mental health indicators. Based on DataCamp dataset.
• Implemented advanced SQL queries analyzing relationships between language proficiency, length of stay, and mental health outcomes.
• Added the 003_llms project, demonstrating how to fine-tune a Llama 2 model using Parameter-Efficient Fine-Tuning (PEFT) with LoRA.
• Developed scripts for training, evaluation, and comparison of the fine-tuned model against its base version.
• Authored a comprehensive guide (llm_finetuning_guide.md) on the step-by-step process of fine-tuning LLMs.

Projects

003_llms: LLM Fine-Tuning with LoRA

This project demonstrates the process of fine-tuning a Llama 2 language model using Parameter-Efficient Fine-Tuning (PEFT), specifically with Low-Rank Adaptation (LoRA). It includes scripts for fine-tuning, evaluation, and comparison, along with a detailed guide on the methodology.

Key Activities:

• Fine-tuning a meta-llama/Llama-2-7b-hf model on a custom JSON dataset.
• Utilizing LoRA for efficient training by adding adapter layers instead of training the full model.
• Scripts to evaluate the fine-tuned model's performance and compare its outputs against the base model.
• A comprehensive guide (llm_finetuning_guide.md) explaining the concepts from data preparation to model inference.

Technologies Used: Python, PyTorch, Hugging Face (transformers, peft, trl, datasets)

006_Tuberculosis_Detection: Computer Vision for Medical Imaging (In Progress)

This project focuses on building a deep learning model to detect signs of tuberculosis in chest X-ray images. It serves as a practical application of computer vision for a real-world medical problem.

Key Activities:

• Baseline Model: A Convolutional Neural Network (CNN) was built from scratch to establish a performance baseline.
• Comparative Modeling: Systematically tested and evaluated multiple architectures (Custom CNN, MobileNetV2, ResNet50) and techniques (Transfer Learning, Fine-Tuning) to diagnose performance issues.
• Data Preprocessing: Implemented a robust data pipeline using ImageDataGenerator for normalization and data augmentation.
• Comprehensive Evaluation: Analyzed model performance not just with accuracy, but with professional diagnostic metrics like Sensitivity, Specificity, PPV, NPV, and the ROC/AUC score.

Technologies Used: Python, TensorFlow, Keras, NumPy, Matplotlib, Scikit-learn

000_Assignments: NLP Auto-Complete

This project involves building a prototype of an auto-complete system using N-gram language models. It's an assignment from Coursera's "Natural Language Processing with Probabilistic Models" course.

Key Features:

• Text preprocessing and tokenization.
• N-gram counting and probability estimation.
• Perplexity calculation for model evaluation.

Technologies Used: Python, NLTK, Pandas, NumPy

001_Data_Analysis: Roller Coaster, Netflix, FRED & Student Mental Health EDA

An exploratory data analysis (EDA) of four datasets: roller coasters, Netflix movies, FRED economic data, and student mental health data. This project involves data cleaning, preparation, visualization, and SQL analysis to uncover insights.

Key Projects:

• Student Mental Health Analysis (SQL): A comprehensive analysis using PostgreSQL to examine the relationship between international students' length of stay and mental health indicators (depression, social connectedness, acculturative stress). Based on DataCamp dataset.
• FRED Economic Data Analysis: Analysis of economic indicators from the Federal Reserve database.
• Netflix Movie Analysis: Exploration of Netflix content library characteristics.
• Roller Coaster Analysis: Statistical analysis of roller coaster features and trends.

Key Findings (Student Mental Health):

• Language proficiency significantly impacts depression levels among international students.
• Length of stay shows varying correlations with different mental health metrics.
• Academic level and age groups display distinct mental health patterns.
• International students face unique challenges compared to domestic students.

Key Findings (FRED Economic Data):

• The S&P 500 index shows significant growth over time, with notable fluctuations during economic events.
• The national unemployment rate exhibits cyclical patterns, with sharp increases during recessions.
• State-level unemployment data reveals regional disparities in economic performance, particularly during the COVID-19 pandemic.

Technologies Used: Python, Pandas, Matplotlib, Seaborn, fredapi, Plotly, PostgreSQL, SQLAlchemy

002_NLP: Natural Language Processing Projects (In Progress)

This folder contains various Natural Language Processing (NLP) projects, including a voice assistant and a sentiment analysis model.

Key Projects:

• IMDb Sentiment Analysis with TensorFlow: A Jupyter Notebook demonstrating sentiment analysis on movie reviews, covering data preprocessing, model building, training, and evaluation.
• Python Voice Assistant: A simple voice-controlled assistant capable of recognizing voice commands and performing basic tasks.

Technologies Used: Python, TensorFlow, Keras, NumPy, Pandas, Matplotlib, Seaborn, SpeechRecognition, pyttsx3, PyAudio

Getting Started

To run these projects, clone the repository and install the required dependencies for each project as listed in their respective README files.

git clone https://github.com/your-username/AI-Learning-Journey.git
cd AI-Learning-Journey

Contact

For any questions or collaborations, please feel free to reach out.