Projects
A selection of platforms, tools, and systems I've architected and shipped.
Telecom Churn Predictor
Binary churn classifier built with scikit-learn (Random Forest, Logistic Regression) and Keras (deep neural network), evaluated against a real-world telecoms dataset, and deployed as a SageMaker real-time endpoint. Includes full model comparison pipeline and Jupyter exploratory analysis.
MLOps Pipeline — SageMaker
End-to-end MLOps pipeline on AWS SageMaker: automated preprocessing, training, evaluation, conditional model registration, and Lambda-triggered retraining on drift detection. Built with the SageMaker Pipelines SDK and Terraform infrastructure.
Real-Time IoT Telemetry Pipeline
Real-time IoT telemetry pipeline for telecoms infrastructure monitoring: multi-device simulator publishes sensor readings to Kinesis, a Lambda consumer writes to DynamoDB with partial batch failure handling, and an async alert handler fires SNS notifications and CloudWatch metrics on threshold breaches.
AI Governance Platform — EKS
FastAPI inference service deployed on Amazon EKS demonstrating production AI governance: per-request audit logging, machine-readable model card with EU AI Act classification and fairness metrics, Helm chart with horizontal pod autoscaling, and Terraform-provisioned EKS cluster with IRSA.