I'm a data enthusiast transitioning into data science, with a background spanning clinical research, public health trials, and digital banking operations. Across every role, data was always at the centre — and that consistency led me to deepen my technical skills in Python, SQL, TensorFlow/Keras, NLP libraries, and Tableau.
I'm passionate about turning raw data into actionable insights, and I build end-to-end projects that reflect real-world analytical workflows across healthcare and banking domains.
End-to-end data science pipeline on 5,000 ClinicalTrials.gov records — EDA, NLP, XGBoost, SHAP, and SQL analytics.
This is my most comprehensive project, combining my clinical research background with a full data science workflow:
| Notebook | Focus |
|---|---|
| 📊 EDA & Data Acquisition | HuggingFace streaming, XML parsing, feature engineering |
| 📝 NLP & Text Analytics | TF-IDF, Sentence Transformers, BART zero-shot, NER |
| 🤖 Machine Learning | XGBoost + hyperparameter tuning (ROC-AUC: 0.68) |
| 🗄️ SQL Analytics | SQLite, window functions, multi-CTE sponsor scorecard |
Key results: Predicted clinical trial completion from registration metadata alone; SHAP explainability identified phase and collaborator presence as the strongest completion signals. NLP baseline (TF-IDF) achieved ROC-AUC of 0.69 from free-text summaries alone.
Tech: Python · XGBoost · SHAP · HuggingFace Transformers · Sentence Transformers · SQLite · Plotly · scikit-learn
End-to-end SQL case study on retail banking operations — customer segmentation, transaction analysis, complaint SLA performance, and an executive KPI scorecard. Built on a synthetic Malaysian banking dataset.
A business intelligence project demonstrating how SQL alone can power executive-grade analytics across a full retail banking operation:
| Query File | Focus |
|---|---|
| 📋 Customer Segmentation | Segment distribution, age bands, tenure cohorts |
| 💳 Transaction Analysis | Monthly trends, MoM change, channel mix, risk flags |
| 📞 Complaints & SLA | SLA breach rates, CSAT scoring, customer risk scorecard |
| 🔄 Cohort Retention | Cohort analysis, dormancy detection, Pareto analysis |
| 📦 Product Performance | Adoption rates, complaint ratios, segment affinity |
| 🏆 Executive Scorecard | 5-CTE composite KPI scorecard across all dimensions |
Key results: Identified cross-sell opportunities in the Mass segment (55.5% of customers, avg 1.91 products); confirmed Pareto — top 20% of customers drive ~58% of transaction value; built a composite risk scorecard combining complaints, SLA exposure, and transaction failure rates.
Tech: SQL · SQLite · CTEs · Window Functions · Python · Pandas · Matplotlib · Seaborn
- Cat vs Dog CNN Image Classifier — End-to-end CNN with TensorFlow/Keras for binary image classification, including data augmentation and evaluation.
- Bank Customer Churn Prediction (ANN) — Artificial Neural Network predicting customer churn for retention strategy insights.
- Vehicle Market Segmentation — K-Means & Hierarchical Clustering to segment vehicles by specification.
- Startup Profit Prediction — Compared multiple regression models to predict startup profitability.
- Drug Classification — Classification models for predicting drug effectiveness.
- Employee Attrition & Retention Analysis — End-to-end HR analytics with EDA, statistical analysis, Tableau dashboards, and predictive modelling.
- Customer Sentiment Analysis — NLP project on hotel reviews using VADER, TF-IDF, and neural networks.
- Python Data Analysis Project — Exploring datasets with Python to uncover patterns and trends.
- SQL Data Analysis Project — SQL-based analysis on real-world datasets for actionable insights.
- Tableau Dashboards — Interactive Tableau dashboards visualizing sales performance.
- Seattle Airbnb Analysis — Analyzing Airbnb listings and pricing trends.
Languages & Tools: Python · SQL · TensorFlow · Keras · Tableau · NLTK · Pandas · NumPy · Scikit-learn · XGBoost · SHAP · HuggingFace Transformers · Matplotlib · Seaborn · Plotly · SQLite · WordCloud
Techniques: Regression · Classification · Clustering · Artificial Neural Networks · Convolutional Neural Networks · NLP · Sentiment Analysis · SHAP Explainability · Data Cleaning · Data Visualization · ETL Pipelines
📌 All projects are fully reproducible with notebooks and environment files included. Explore my repositories for the full workflow.