Jayanth Kumar

·         Designed and deployed scalable ML pipelines, orchestrating data ingestion from EMR, operational systems, managing model training, validation, deployment, automated retraining; enabled continuous production updates, reduced manual intervention, ensured reproducibility, audit readiness. ·         Built scalable LLM orchestration frameworks using LangChain, LangGraph, and CrewAI to support multi-agent enterprise workflows and autonomous decision-making systems. ·         Developed scalable AI/ML backend services using Python, FastAPI, and Flask for enterprise-grade Agentic AI and RAG-based applications. ·         Developed predictive patient risk models using XGBoost, LightGBM, PyTorch, TensorFlow 2.x, combining structured clinical records, lab results, temporal vitals data; applied ensemble methods, LSTM-based architectures to improve early intervention accuracy ·         Implemented secure and scalable LLM deployment architectures on Kubernetes and Docker using GPU-enabled infrastructure across AWS, Azure, and GCP environments. Developed scalable SQL-based ETL pipelines for ingesting, transforming, and validating structured and semi-structured enterprise data from multiple source systems.

·         Designed, implemented, and deployed ML pipelines, automating data ingestion from multiple insurance sources, feature engineering, delivering scalable production deployments that supported real-time, batch scoring for risk, fraud, customer segmentation use cases. ·         Designed and managed an enterprise feature store, implementing version-controlled feature pipelines, automated refresh schedules, and cross-team access policies, enabling consistent feature reuse across projects and reducing model development time. ·         Automated CI/CD workflows and Kubernetes deployment operations using Python scripting across AWS, Azure, and GCP environments. ·         Implemented GraphQL subscriptions and event-driven APIs for real-time monitoring, notifications, and AI workflow orchestration systems. ·         Designed relational database schemas and normalization strategies using SQL for scalable and secure enterprise applications. ·         Built and deployed production-grade risk prediction, fraud detection, customer segmentation models, incorporating insurance-specific features, achieving a measurable lift in prediction accuracy, supporting automated underwriting, fraud detection workflows. ·         Integrated transformer-based NLP models with Hugging Face Transformers to automate claims document parsing, reducing manual review by 60% and improving data extraction accuracy for underwriting and adjudication workflows. ·         Optimized AstraDB vector indexing strategies to improve embedding search accuracy, query performance, and response latency across large datasets. ·         Implemented REST API authentication and authorization using OAuth2, JWT, API Gateway, and RBAC security models. ·         Implemented GraphQL APIs and webhook-based integrations to streamline communication between AI agents and enterprise applications.

Built real-time fraud detection and credit risk scoring systems across card payments and online banking, with full PCI-DSS and SOX compliance. - Developed real-time fraud detection pipelines processing card and online banking transactions, triggering instant alerts for risk teams - Built time-series forecasting models for transaction volumes, enabling proactive cash replenishment and capacity planning - Designed explainable AI dashboards using SHAP and LIME to support regulatory compliance and audit readiness - Implemented NLP models to categorize and route customer inquiries, reducing manual support handling - Mentored junior data scientists on feature engineering, model validation, and MLOps best practices

Delivered ML and NLP solutions for enterprise clients across retail and financial services, focusing on fraud detection, forecasting, and document intelligence. - Designed NLP pipelines for entity extraction, document classification, and sentiment analysis, reducing manual processing overhead - Built and monitored production ML models for fraud detection and risk prediction across high-volume transactional datasets - Containerized and deployed ML APIs with Docker and Flask on AWS EC2, ensuring reproducible client-ready environments - Implemented drift monitoring and automated retraining triggers to maintain model performance across client deployments

- Developed revenue forecasting models using Linear Regression and Random Forest to improve audit materiality assessments - Automated data ingestion and preprocessing workflows, reducing manual effort and accelerating data availability - Designed interactive dashboards in Tableau and QlikView to visualize revenue trends and forecast accuracy for stakeholders