Senior AI/ML Engineer (GenAI/NLP) at Capital One

Designed and deployed an enterprise-scale Retrieval-Augmented Generation (RAG) system to automate financial knowledge retrieval across multiple internal datasets, reducing manual analyst effort by 40%. Fine-tuned LLaMA2 70B models on proprietary transaction and policy data to optimize embeddings for real-time retrieval. Built hybrid semantic search pipelines using FAISS and Elasticsearch for sparse-dense retrieval with sub-second latency across millions of records. Engineered a dynamic prompt orchestration framework with LangChain for template versioning and traceability across LLM endpoints. Containerized inference microservices with BentoML, Ray Serve, and Docker to support high-availability REST and gRPC endpoints. Automated distributed training on Kubernetes (EKS) with GPU scheduling and elastic scaling. Implemented model observability dashboards with Prometheus and Grafana; integrated drift and throughput monitoring and alerting. Developed explainability pipelines with SHAP and cu

AI/ML Engineer (Healthcare AI / Predictive Analytics) at Molina Healthcare

Developed HIPAA-compliant pipelines for predictive patient risk scoring across claims, clinical records, and engagement data to support preventive care targeting. Implemented structured FHIR/HL7 data ingestion and transformation to ensure standardized, reliable training datasets. Built TensorFlow 2.10 models for patient readmission and adverse event likelihood, enabling more efficient clinical workflows. Designed automated feature engineering on Google Cloud Dataflow for scalable real-time risk scoring. Deployed production inference services on Vertex AI for low-latency analytics dashboards. Produced explainable AI outputs with SHAP and LIME to support care teams and compliance audits. Implemented Terraform-based CI/CD and Cloud Build to automate model and preprocessing deployment. Established retrieval of data quality and drift detection via Airflow 2.6 with triggers based on population changes and metrics. Enforced RBAC for sensitive patient data with audit trails and multi-team acce

Senior Data Scientist / ML Engineer at Lumen Technologies

Developed end-to-end predictive maintenance models using historical performance and event logs, reducing downtime by 30%. Built graph-based network analysis with PyTorch Geometric to identify high-risk nodes and optimize routing. Designed streaming anomaly detection with Apache Kafka and Spark Structured Streaming for real-time incident detection. Deployed containerized ML services with KServe for production-grade inference and high-throughput endpoints. Automated feature extraction from multi-source logs with Apache Beam to produce reusable features. Built a time-series forecasting pipeline with Prophet and XGBoost to forecast bandwidth utilization for dynamic resource allocation. Implemented model observability dashboards with Grafana and Prometheus to track drift, latency, and accuracy. Performed hyperparameter optimization with Optuna, achieving ~12% higher accuracy on network failure events. Created automated model promotion pipelines for staging/production governance. Developed i

Data Scientist / ML Engineer (Retail & E-commerce AI) at Kohl’s

Developed end-to-end customer behavior analytics pipelines leveraging transaction and loyalty data to optimize personalized marketing campaigns. Built recommender systems using LightGBM and Word2Vec embeddings, increasing upsell conversions by 18%. Designed image-based product classification with TensorFlow to automate catalog tagging and support visual search. Implemented batch and real-time feature pipelines with Airflow for consistent features across models. Built inventory demand forecasting with LSTM to reduce overstock and stockouts by 15%. Created automated model evaluation and A/B testing pipelines to measure performance. Deployed microservices for real-time inference of customer scoring and personalization with FastAPI. Created visualization dashboards with Seaborn and Plotly to monitor KPIs. Standardized pipelines for unifying disparate product metadata and ensured clean input to models. Implemented automated SME labeling corrections to improve subsequent retraining accuracy.