Summary
Lilong Jiang is a Principal Machine Learning Engineer with 12 years of experience and a Ph.D. in Computer Science, now leading core recommendation and retrieval modeling at Roblox from Redwood City. He has a proven track record of productionizing large-scale ML systems—reducing training time and cost by over 90% with IterableDataset, migrating retrieval and ranking stacks from TensorFlow to PyTorch, and building a custom two-tower PyTorch library with in-batch negative sampling and bias correction. Previously at Twitter he led notification relevance and candidate generation efforts that boosted engagement by millions of daily users and drove cross-team migrations and heavy-ranker initiatives. His strengths span advanced deep-learning architectures (DIN/DIEN, transformers, DCNv2), user-history modeling, and practical debiasing techniques that tangibly improved bookings and playtime. Colleagues describe him as a tech lead who combines research rigor from his academic background with pragmatic engineering to ship measurable product impact. An under-the-radar detail: he repeatedly bridges retrieval and ranking gaps end-to-end—designing sampling and correction methods as well as scalable training pipelines.
12 years of coding experience