Case Study: PAICON accelerates oncology AI research and cuts training costs with Vast.ai

PAICON cuts research-phase training costs by 60%+ with Vast.ai

Paicon, a global oncology data platform, needed to rapidly iterate on its Athena histopathology foundation model during the research phase, requiring extensive parallel experimentation with substantial GPU compute. Using hyperscale cloud providers for this high-volume R&D work proved economically challenging due to premium pricing, so they sought a more cost-efficient solution to enable faster iteration without prohibitive costs.

Paicon implemented a hybrid compute solution using Vast.ai's GPU cloud for research-scale experimentation. This allowed the team to run parallel trials on multi-GPU machines with on-demand scaling. By shifting large training workloads to Vast.ai, Paicon achieved a research-phase training cost reduction of over 60% and gained greater experimental agility, accelerating their development cycles and validating their diversity-aware training approach.

Paicon