Senior AI Engineer – GenAI, LLM & Agentic Systems at UnitedHealth Group

Led enterprise GenAI initiatives using LangGraph and Azure services to coordinate multi-agent workflows for crawling and synthesizing Electronic Medical Records (EMRs). Built GraphRAG pipelines using Azure OpenAI (GPT-4o) and Azure AI Search to transform longitudinal patient histories into structured, entity-linked knowledge graphs, enabling faster semantic search and retrieval for clinical teams. Produced explainable, chronologically indexed patient portfolios stored in Azure Cosmos DB with source-cited summaries for clinical review. Implemented tool/data access standardization via Model Context Protocol (MCP) servers, exposing Azure AI Search, Cosmos DB, and internal clinical APIs as unified context providers across agent applications. Developed secure RAG ingestion/ETL using Azure Data Factory and Azure Functions into vector stores. Designed entity resolution models with Azure ML to unify medical term and ICD/medication variations into consistent patient profiles. Built Human-in-th

Senior AI ML Engineer at Fifth Third Bank

Scaled and optimized the Jeanie chatbot platform by refining NLU models to achieve ~90% intent recognition for high-volume banking queries. Hosted and managed conversational models on AWS SageMaker with custom Docker containers for sub-second real-time inference. Piloted CrewAI agent crews for call-transcript triage (sentiment scoring, PII redaction, intent classification) ahead of LangGraph adoption. Built mobile app navigation prediction pipelines using Transformer models deployed on AWS EKS. Developed NLP pipelines ingesting and analyzing contact center calls for real-time sentiment tracking and intent mapping. Implemented serverless audio ingestion pipelines using S3, Lambda, and Transcribe; applied Comprehend and custom spaCy models for NER/PII masking and sentiment scoring. Established reusable AWS ML pipelines with centralized feature stores (SageMaker Feature Store), built RAG evaluation baselines with RAGAS, and created automated MLOps for training/evaluation/registry promoti

Machine Learning Engineer at Kroger

Built time-series forecasting pipelines for daily inventory demand using Prophet, LightGBM, and XGBoost to reduce waste while improving product availability. Engineered distributed ETL with Apache Spark (PySpark) to clean and transform POS data for supply-chain modeling, including rolling-window validation and hyperparameter tuning across thousands of SKUs. Developed reinforcement learning and optimization heuristics for logistics routing to reduce fuel consumption and delivery windows, including MDP-based route simulation with real-time traffic updates. Collaborated on edge-deployed computer vision pipelines for automated checkout loss prevention: trained CNNs and deployed object detection models (YOLO/Faster R-CNN) on NVIDIA Jetson with TensorRT for low-latency inference. Set up annotation pipelines for continuously labeling edge-case images, built Kafka streaming to route out-of-stock events to mobile apps, and monitored edge-device health and model degradation. Used containerized

Machine Learning engineer at AT&T

Built predictive pipelines using scikit-learn to identify high-risk telecom subscribers for retention, leveraging billing history and network usage metrics. Created feature selection pipelines using variance thresholding and RFE. Addressed class imbalance using SMOTE and cost-sensitive learning. Developed supervised time-series forecasting and geospatial analytics for high-demand cellular zone identification, including spatial indexing (H3/Geohash) and capacity strain modeling with ARIMA/SARIMAX. Implemented large-scale ETL with Apache Spark (PySpark) and Hadoop/Hive to process multi-terabyte data lakes, optimizing Spark runtimes by reducing shuffle bottlenecks and tuning partitions. Structured data into columnar formats (Parquet/ORC) and orchestrated workflows with Airflow DAGs. Packaged reusable preprocessing pipelines in Docker and added schema/feature stability tests with PyTest. Coordinated version control and CI/CD with Git to deploy verified updates across staging/production, f