Designed Al-driven financial forecasting models using Auto-Sklearn and H2O.ai for market trend prediction and risk estimation, improving forecast accuracy by 15%.

Built fraud detection pipelines on AWS SageMaker and PySpark with Kinesis Data Streams; improved true positive rates by 12% and reduced manual case reviews by 20%.

Engineered advanced fraud risk features (velocity thresholds, geolocation mismatches, device fingerprinting), lowering false positive alerts by 18%.

Developed machine learning models for programmatic ad bidding optimization using Python, TensorFlow 2.x, PyTorch 1.13, and XGBoost on AWS SageMaker, Kubernetes, with real-time scoring supported by Lambda and Kinesis.

Led full cloud migration of ML workflows from AWS SageMaker to GCP Vertex Al, unifying training, deployment, and monitoring pipelines, reducing bid cost variance by 15% and accelerating deployments.

Collaborated with infra and business teams to align Agentic Al workflows (planner-executor orchestration, tool calling, guardrails) with compliance, marketing, and fraud detection; integrated Jira, Confluence, and Git for traceable and auditable workflows.

Led end-to-end development of AI/ML solutions for personalized recommendations, targeted promotions, and supply chain optimization, driving measurable ROI in customer engagement and operational efficiency.

Architected and implemented high-throughput ingestion pipelines with AWS Glue, Spark, PySpark, and SQL-based preprocessing to process clickstream, POS, and inventory data, reducing pipeline latency by 25%.

Built and deployed containerized microservices (Django, Tornado, FastAPI with async/await, Docker) on AWS EKS, delivering low-latency APIs for recommendation engines and pricing intelligence.

Partnered with fraud prevention and risk analytics teams to design ML models for real-time transaction scoring, reducing false positives and improving fraud detection accuracy.

Built and validated classification and anomaly detection models (Logistic Regression, Random Forests, XGBoost, K-means) for fraud detection and merchant risk scoring.

Developed deep learning NLP models (Keras, PyTorch) for text classification, entity extraction, and summarization, automating dispute resolution and compliance reviews.

Contributed to end-to-end analytics projects, from requirements through data ingestion, transformation, modeling, and visualization for enterprise clients.

Designed and maintained large-scale pipelines with Hadoop HDFS, Hive, Spark, supporting structured/unstructured datasets.

Performed behavioral clustering using K-means on customer datasets to uncover actionable insights and drive targeted marketing strategies.