AI/ML Engineer at Principal Financial Group

Fine-tuned LLMs (Llama 2, Mistral) using SFT and RLHF with distributed training, improving perplexity and downstream accuracy. Architected real-time fraud screening pipelines using AWS, Python/PySpark, and Kafka with Docker/Kubernetes anomaly-detection microservices. Productionized credit-risk models (XGBoost, Logistic Regression) via MLflow/FastAPI into lending decision engines. Built RAG-powered LLM copilots (BERT/T5, Azure OpenAI, vector DBs) adopted by risk teams. Automated end-to-end ML pipelines with MLflow and CI/CD to reduce release cycles. Deployed and optimized 20+ containerized ML/LLM models on AWS as real-time REST/gRPC endpoints with sub-50ms p95 latency; reduced GPU inference costs via quantization and batching. Mentored junior data scientists on MLOps and deployment workflows.