Summary
Jefferson Pinares is a Machine Learning Engineer and educator with over a decade in tech and more than six years deploying AI systems in production across NLP, Speech AI, Computer Vision, and LLMs. He has driven end-to-end, cloud-native GenAI solutions with strong MLOps, CI/CD, and observability practices, and led voice AI efforts that produced production TTS with near-zero WER and halved operational costs versus major cloud providers. His hands-on work spans LLM fine-tuning, RAG, conversational agents, real-time voice integration, model acceleration (ONNX, compilers) and Kubernetes-based scaling, enabling multimodal products used across 20+ countries. An MSc scholar with publications and talks at NeurIPS, NAACL, ICML and Khipu, he pairs research credibility with measurable product impact. Jefferson also trains the next generation of AI engineers in Latin America, blending practical deployment experience with academic rigor. Less obvious: he has repeatedly shortened ML release cycles (up to 60%) and slashed data prep time by 70% through automated pipelines and tooling.
10 years of coding experience
10 years of employment as a software developer
Curso de Especialización, Database Management, Curso de Especialización, Database Management at Universidad Nacional Mayor de San Marcos
Deep Learning - Reinforcement Learning Summer School, Deep Learning - Reinforcement Learning Summer School at Mila - Quebec Artificial Intelligence Institute
MicroMaster, Statistics and Data Science, MicroMaster, Statistics and Data Science at Massachusetts Institute of Technology
Diplomado, Data and Information Management, Diplomado, Data and Information Management at Universidad Nacional San Antonio Abad del Cusco
Master of Science - MS, Computer Science - Artificial Intelligence, Master of Science - MS, Computer Science - Artificial Intelligence at Universidad Católica San Pablo
Khipu Summer School en Artificial Intelligence, Khipu Summer School en Artificial Intelligence at Universidad de la República
Spanish, English, Portuguese