Machine Learning Projects Built for Real-World Impact
This portfolio showcases end-to-end machine learning and data science projects. Each project demonstrates structured problem solving, rigorous data preparation, robust model evaluation, and a strong focus on explainable and reliable models. I design solutions that translate complex data into practical, decision-support tools with measurable impact.
Diabetes Risk Prediction with Explainable AI
Diabetes Risk Prediction with Interpretable Machine Learning
Developed supervised machine learning models to predict diabetes risk using clinical patient data. Built end-to-end pipelines in Python (pandas, scikit-learn, NumPy), including data cleaning, feature engineering, and cross-validation.
Integrated SHAP to provide feature-level explanations and subgroup analysis, ensuring transparency in healthcare decision support. Delivered a Streamlit dashboard for model inspection and practical deployment.
Key skills:
Python, scikit-learn, SHAP, cross-validation, model evaluation, healthcare analytics, decision support.
Explainable Loan Approval Decision System
Explainable Loan Approval System (XGBoost + SHAP + Policy Logic)
Designed a decision-support system combining gradient boosting models with SHAP-based interpretability and policy-grounded validation.
Implemented structured feature engineering, SQL-based data preparation, and rigorous performance evaluation. The system provides defensible approval/rejection explanations aligned with business rules and compliance requirements.
Key skills:
XGBoost, SHAP, SQL, model validation, error analysis, explainable AI, decision support systems.
LLM Document Intelligence with RAG
LLM-Powered Document Intelligence (RAG System)
Designed and implemented a Retrieval-Augmented Generation pipeline for structured document analysis. Built ingestion, chunking, embedding, and semantic retrieval workflows in Python.
Developed evaluation layers for groundedness, hallucination reduction, and response traceability. Structured modular components suitable for version control and auditability.
Key skills:
RAG, embeddings, LangChain-style workflows, LLM evaluation, groundedness checks, traceability, prompt engineering.
Pharmaceutical Property Prediction
Machine Learning for Pharmaceutical Property Prediction
Built and validated predictive models for pharmaceutical research using noisy experimental datasets. Designed robust feature engineering pipelines and performed cross-validation, error analysis, and stability testing.
Documented reproducible ML workflows for regulated research environments.
Key skills:
Python, pandas, scikit-learn, robustness testing, reproducibility, regulated ML workflows.
Create Your Own Website With Webador