Joseph Rivas

About

ML/AI engineer with hands-on ownership of production document intelligence and GenAI systems, spanning model experimentation, AWS deployment, monitoring, and iterative optimization. Stands out for turning document-heavy workflows into reliable, near real-time products with measurable gains in accuracy, latency, and manual-effort reduction, while also shipping citation-grounded RAG features that drove user trust and adoption.

Key Strengths

• Owned ML systems end-to-end from experimentation through production monitoring
• Built document intelligence pipeline for low-quality, inconsistent PDFs
• Improved effective field-level accuracy to 90% on critical fields
• Reduced manual document processing time by 50%
• Delivered near real-time processing for most document batches
• Strong production ML architecture on AWS with SageMaker, Lambda, API Gateway, S3, and CloudWatch
• Applied human-in-the-loop design to improve reliability on low-confidence predictions
• Shipped RAG-based document Q&A and summarization grounded in client-specific documents
• Improved user trust through citation-grounded GenAI responses
• Balanced LLM quality, latency, and cost with chunking, context limits, and caching
• Reduced latency by 30% through production optimization
• Improved answer precision with reranking and retrieval refinements
• Built modular Python ML services with FastAPI and Flask
• Established structured testing with pytest, integration tests, schema validation, and golden datasets

Experience

Education

Skills

Languages

Publications

2 publications