Integrated autonomous agent workflows into clustering and anomaly detection pipelines, automating feature extraction and reducing manual ML overhead across Retail and Media clients. • Automated legacy Marketing Mix Models via Python and Bayesian optimization, engineering an end-to-end ML pipeline that cut model turnaround from 45 days to 7 days (85% reduction). • Built scalable Spark + Airflow E2E ML pipelines for US-based Retail and Banking clients, improving data ingestion reliability and reducing feature engineering latency across production workflows. • Applied hyperparameter tuning via Bayesian optimization and statistical validation to improve model accuracy and interpretability across deployed campaign decision-making systems