infant-joshva/data-science-learning-path

Purpose and Scope

This is a data-science learning portfolio, not one deployable application. It organizes curriculum notebooks from Python/SQL through ML, deep learning, NLP, RAG, computer vision, AWS, and five standalone analytics/ML projects. The root overview is README.md; curriculum material begins at DS-Practice-Notebooks/README.md.

Execution Entry Points

Jupyter notebooks are the principal execution and training entry points, including DS-Practice-Notebooks/DS03-eda-casestudy-ml/EDA_Git.ipynb and ML Project-CropYieldPredictor/notebook/Crop_Production_Prediction.ipynb. Runnable Python application entry points exist for the Employee Attrition, Multiple Disease, and sentiment projects: Project_3(Employee_Attrition_ML_Model)Git/app/main.py, Project_4(Multiple_Disease_Prediction)/app/main.py, and Project_5(AI Echo - Sentiment Analysis)/app/main.py.

Learning-to-Project Structure

The curriculum is staged by capability:

• Python, SQL, Pandas data handling: DS-Practice-Notebooks/DS01-python-sql-pandas/Py-SQL-Pandas_Git.ipynb.

• Statistics and scientific computing: DS-Practice-Notebooks/DS02-stats-numpy-scipy/Stats_NumPy_&_SciPy_Git.ipynb.

• EDA, case studies, and supervised ML: DS-Practice-Notebooks/DS03-eda-casestudy-ml/ML/1_Supervised Learning/ML_Hands_on.ipynb.

• Deep learning, NLP, computer vision, RAG, and AWS: DS-Practice-Notebooks/DS04-DL-NLP/RAG/Document_RAG.ipynb, DS-Practice-Notebooks/DS04-aws/CustomLayerAWS.ipynb.

Data and Model Lifecycle

Projects generally follow a local artifact flow: raw CSV data → notebook/ETL-based cleaning and feature engineering → cleaned CSV → model training → serialized .pkl model → application inference. This boundary is explicit in the disease project: Project_4(Multiple_Disease_Prediction)/data/raw_dataset, Project_4(Multiple_Disease_Prediction)/data/cleaned_dataset, and Project_4(Multiple_Disease_Prediction)/models. Training and serving are decoupled; apps consume already-trained artifacts rather than retraining at request time.

Analytics and Persistence Boundary

Two projects include PostgreSQL-oriented ETL scripts, separating source-file preparation from reporting or downstream querying: Project_2(AgriData_Explorer)Git/agri_etl_postgres.py and Project_3(Employee_Attrition_ML_Model)Git/attrition_etl_postgres.py. Power BI files provide a separate reporting/presentation layer: Project_2(AgriData_Explorer)Git/Agri_dashboard.pbix and Project_3(Employee_Attrition_ML_Model)Git/dashboard/Attrition report.pbix.

Deployed ML Project Families

Employee attrition supports model selection and alternate imbalance-aware pipelines, with persisted logistic-regression/SMOTE and SVC/ADASYN models in Project_3(Employee_Attrition_ML_Model)Git/models. Multiple disease prediction routes inference across kidney, liver, and Parkinson’s artifacts in Project_4(Multiple_Disease_Prediction)/models. AI Echo performs review sentiment inference from versioned Naive Bayes and logistic-model artifacts in Project_5(AI Echo - Sentiment Analysis)/model.

Material Runtimes and External Assets

The repository is primarily Python, executed in Jupyter and app runtimes; dependencies are declared globally in requirements.txt and project-specifically in Project_3(Employee_Attrition_ML_Model)Git/requirements.txt. Core data/ML tooling includes Pandas, NumPy, scikit-learn, TensorFlow/Keras, and visualization libraries per the root README. External or binary assets include Roboflow/YOLO-focused notebooks and a trained PyTorch model at DS-Practice-Notebooks/DS04-DL-NLP/Computer_vision/model/best.pt, plus local RAG input DS-Practice-Notebooks/DS04-DL-NLP/RAG/sample.pdf.