Summary
Borislav Mavrin is a Senior Machine Learning Engineer with nine years of experience building production-grade AI systems across consumer tech and enterprise environments. He has led teams and research at LG and Huawei, delivering state-of-the-art multimodal QA, efficient Fusion-in-Decoder RAG systems, and published RL research with a US patent. At Coveo he focused on agent-based search, automated prompt tuning, and robust evaluation for retrieval-augmented generation pipelines, and he recently reproduced GPT-OSS-20B SWE-Verified results with a native Harmony Agent. Comfortable shifting between research and MLOps, he optimizes memory, latency, and deployment (PyTorch, FAISS, Docker, distributed training) to bring models from paper to product. Based in Edmonton with deep academic roots (University of Alberta, Concordia) and industry collaborations (University of Toronto), he combines rigorous statistical ML training with fast iteration in applied conversational AI. An uncommon strength is scaling complex retrieval+LLM systems by orders of magnitude in footprint and latency while preserving accuracy.
9 years of coding experience
9 years of employment as a software developer
Master's degree Statistical Machine Learning, Master's degree Statistical Machine Learning at University of Alberta
Deep Learning Summer School 2016 Deep Learning, Deep Learning Summer School 2016 Deep Learning at Université de Montréal
Master's Degree Mathematics, Master's Degree Mathematics at Concordia University
French, Russian, English