AI/ML Engineer at USAA

Developed and deployed LLM pipelines using LangGraph, LlamaIndex, and GPT-4 to automate document intelligence for financial compliance reports, reducing manual review time by 42%. Engineered a Retrieval-Augmented Generation (RAG) system with Pinecone and FAISS to retrieve unstructured credit data with under 1-second latency, improving analyst decision accuracy by 31%. Trained and fine-tuned transformer models (BERT, LLaMA, Claude) for NER with 96% precision. Implemented scalable TensorFlow pipelines on AWS SageMaker for customer risk scoring, reducing training costs by 28%. Automated data ingestion and preprocessing using PySpark and PostgreSQL, enabling near real-time analytics for fraud detection and doubling pipeline throughput. Orchestrated containerized deployments via Docker, AWS Lambda, and CI/CD workflows, ensuring consistent and version-controlled model releases. Integrated Generative AI models for conversational query handling within internal analytics tools, improving employ