Deepthi Chava

Built a hypothesis-driven root cause analysis (RCA) intelligence layer that correlates logs, metrics, traces, and incident records to produce confidence-scored root cause outputs, reducing MTTR and eliminating manual war-room diagnostics. Implemented agent-based evidence routing, confidence scoring, explainable causal graphs, and governance patterns for reproducible RCA on GCP.

Built scalable GenAI and retrieval-augmented generation (RAG) pipelines to operationalize LLM-powered assistants for auditing automation, claim validation, and healthcare data abstraction. Deployed GPU-efficient RAG microservices with BentoML and KServe on GKE and implemented hybrid retrieval, prompt tracing with PromptLayer, and version-controlled model updates with MLflow and DVC.

Architected secure, explainable, reproducible ML deployment infrastructure for cybersecurity analytics pipelines on Azure with CI/CD-enabled workflows and operational visibility. Containerized and deployed multi-framework inference using TorchServe, TensorFlow Serving, and Triton on AKS, with feature pipelines in Feast and explainability via SHAP and LIME.

Built and scaled ML platform capabilities for financial risk and fraud analytics, delivering reproducible training pipelines, containerized model lifecycle management, and CI/CD automation for Kubernetes deployments. Integrated DVC/MLflow lineage tracking with GitOps approvals and rollback, deployed SageMaker scoring services, and added drift detection gates and Prometheus-based monitoring.

Developed ML-driven customer intelligence systems for churn risk scoring, behavioral segmentation, and campaign optimization in the retail analytics domain. Built supervised and deep learning models with scikit-learn and Keras, including preprocessing, SMOTE-based imbalance handling, PCA/K-means segmentation, and evaluation with GridSearchCV and standard classification metrics.

Developed and maintained ETL pipelines and reporting for enterprise client datasets using Python and SQL. Optimized complex queries and materialized views, implemented SSRS reports and data quality checks, and documented ETL lineage and table mappings for compliance.