Lin Qiao is co-founder and chief executive of Fireworks AI, which builds infrastructure for training, customizing, and serving open-weight models. A former head of PyTorch at Meta, she has spent her career making ambitious computing systems practical, from databases and distributed data platforms to machine-learning frameworks and production AI.
Qiao earned a doctorate in computer science at the University of California, Santa Barbara, and worked on the BLU query engine at IBM’s Almaden Research Lab. At LinkedIn, she helped develop the Espresso distributed data platform and Gobblin data-ingestion framework, eventually leading data lifecycle management for analytics. She joined Meta in 2015, worked on infrastructure for Caffe2 and PyTorch, and coauthored research on deploying inference accelerators.
At Fireworks, Qiao applies that systems background to a central commercial problem: companies want models adapted to their own customers and workflows without assembling GPU orchestration, optimization, deployment, and monitoring infrastructure themselves. Her developer-centric AI philosophy combines accessible tooling with three interconnected priorities:
- Model ownership as strategic independence. Open-weight models give companies greater control over critical capabilities and let them build specialized intelligence around proprietary data instead of depending entirely on outside model providers.
- Training and inference as one system. Fireworks’ integrated training platform connects supervised fine-tuning, reinforcement-based adaptation, deployment, and customized serving, allowing production experience to inform ongoing model improvement.
- Performance as product infrastructure. Latency, reliability, routing, and compute efficiency shape whether AI applications work at commercial scale. Fireworks’ acquisition of Hathora expanded its capacity for distributed compute orchestration and latency-sensitive deployment.
Qiao’s aim is specialized production intelligence: AI systems that companies can shape around their own knowledge, operate reliably, and improve over time.
