Architecture decisions on record, in code
Six repositories across AI/ML, cloud infrastructure, data pipelines, and automation. Stack, structure, and rationale — readable in minutes, not hours.
AI/ML · Cloud · Data · Automation
ai-claims-predictor
cloud-migration-accelerator
customer-360-pipeline
XGBoost pipeline predicting insurance claim outcomes with 87% accuracy. Includes feature engineering, SHAP explainability layer, and FastAPI inference endpoint.
Terraform modules and runbooks for lift-and-shift to AWS. Covers VPC, IAM baseline, RDS migration, and blue-green deployment patterns used in production.
Apache Spark ETL ingesting policy, claims, and CRM data into a unified customer graph. Reduces cross-domain query latency from 4 hours to under 8 minutes.
Python · XGBoost · FastAPI · AWS SageMaker
Python · Spark · Delta Lake · Azure Synapse
Terraform · AWS · Python · GitHub Actions
fraud-detection-engine
workflow-automation-platform
pricing-model-engine
Real-time anomaly detection on transaction streams using Isolation Forest and LSTM. Deployed on Kafka with sub-100ms latency; flags reviewed in a React audit UI.
Event-driven automation engine built on Apache Airflow. Orchestrates underwriting, compliance checks, and document processing — cutting manual handoffs by 60%.
Gradient-boosted pricing model with automated retraining on new loss data. REST API consumed by underwriting systems; full MLflow experiment tracking included.
Python · Kafka · TensorFlow · GCP Vertex AI
Python · Airflow · PostgreSQL · Docker · AWS
Python · LightGBM · MLflow · FastAPI · Azure ML
Every repo includes a README with the decision log
Not just what was built — why those technology choices, what tradeoffs were accepted, and what the production outcome measured. Verifiable in the commit history.