By Renu Prakash
Mentored by James Gearheart, VP & Senior Data Scientist (AI/ML), Wells Fargo
Welcome! This portfolio is a curated collection of hands-on, end-to-end data science and machine learning projects simulating real-world challenges in financial crime detection, risk analytics, and compliance. My work blends AI/ML technical depth, business context, and practical experience to deliver measurable impact across diverse fraud types, modeling challenges, and regulatory domains.
Financial fraud is ever-evolving—exploiting identity, behavioral, and system vulnerabilities. Institutions need robust, explainable solutions. Here, I demonstrate how advanced data science and AI can be practically applied to build scalable, risk-aware fraud detection and compliance systems that meet industry and regulatory expectations.
- End-to-End ML Ownership: Each project covers the full cycle—from data exploration and feature engineering to modeling, evaluation, and interpretability.
- Business-Aligned Impact: Solutions are grounded in real business needs, including fraud, risk, and KYC/AML compliance.
- Breadth & Depth in Data Science: My portfolio spans classic machine learning, unsupervised anomaly detection, deep learning, clustering, and explainable AI.
- Outcome Focused: Projects emphasize measurable results, clear visualizations, and actionable business value.
- Excellent Communication: I present findings and methods in a way that’s accessible for both technical and non-technical audiences.
-
Transaction Fraud Detection:
Built an imbalanced-class pipeline using Random Forest, SMOTE, and SHAP, increasing fraud recall from 6% to 24%.
View Project → -
Deepfake Detection with CNNs:
Developed hybrid deep learning models (image + tabular) for synthetic ID/media detection, achieving 88% accuracy.
View Project → -
Check Kiting Detection using Unsupervised Learning:
Created behavioral features at the customer level and applied Isolation Forest to flag the top 0.5% of high-risk accounts, uncovering patterns consistent with systematic float exploitation in synthetic banking transactions.
View Project → -
Identity Fraud Detection:
Built advanced classifiers using both transactional and identity metadata, ensemble ML, and clustering to catch impersonation and identity theft. Achieved a 41% increase in fraud recall and surfaced high-risk clusters using account velocity and device fingerprinting.
View Project → -
Synthetic Identity Fraud Detection:
Modeled emerging synthetic identity risks using metadata and document analysis, neural networks, and tabular modeling to flag fake or mixed identities.
View Project → -
More Projects:
- Exploratory Data Analysis (EDA) for credit card fraud
- Account takeover (ATO) fraud detection using session and behavioral analytics
- KYC/AML compliance system for regulatory risk monitoring and dashboards
(Browse the repo directory below for details on all projects)
- Programming & Tools: Python, SQL, R, Jupyter, Git, Tableau, Power BI, scikit-learn, TensorFlow, pandas, NumPy, MobileNetV2
- Machine Learning & Analytics: XGBoost, Random Forest, Isolation Forest, Voting Ensemble, PCA, SMOTE, K-Means, EDA, Anomaly Detection, Risk Scoring, Model Monitoring, Convolutional Neural Networks (CNN), SHAP
- Visualization: SHAP-based feature attribution, compliance dashboards, Matplotlib, Seaborn, Plotly, Tableau, ggplot2
- Domain Expertise: Fraud analytics (transaction, identity, synthetic, deepfake, ATO), AML/KYC, Credit Risk, Economic Index Modeling, Financial Forecasting, Business Strategy, ATM Deployment Analytics
- Applied Economics & Statistics: Scenario planning, index numbers, economic indices, applied statistics
- Communication & Leadership: Business communication, strategic reporting, stakeholder engagement, policy development
| Fraud Type | Detection Method | Key Technologies |
|---|---|---|
| Transaction Fraud | Supervised ML, anomaly detection | Random Forest, SMOTE, PCA, SHAP |
| Identity Fraud | Clustering & behavioral modeling | XGBoost, K-Means, account velocity analysis |
| Synthetic Identity Fraud | Metadata & document analysis | Neural nets, tabular modeling |
| Check Kiting | Unsupervised anomaly detection | Isolation Forest, behavioral profiling |
| Deepfakes | Image/audio forensics | CNN, MobileNetV2, TensorFlow |
| Account Takeover (ATO) | Behavioral/session analysis | PCA, Random Forest, time feature extraction |
| AML/KYC Compliance | Integrated risk modeling | Feature attribution, compliance scoring |
- Transaction Fraud Detection
- Deepfake Detection with Neural Network
- Check Kiting Detection using Unsupervised Learning
- Identity Fraud Detection
- Synthetic Identity Fraud Detection
- Account Takeover Fraud Detection
- KYC/AML Compliance System
- Credit Card EDA Analysis
I'm passionate about using AI and data science to solve meaningful problems and deliver real-world impact. I’m open to exploring diverse roles where I can apply my skills across fraud analytics, data science, financial risk, and beyond.
If you're building something impactful—let’s connect, collaborate, and make a difference together.