Eugene Cheah is co-founder and CEO of Featherless, a platform for running open AI models through an API. He also co-leads RWKV, an open model-architecture project.[1][2][3]
Start with The AI Bill Is Too Damn High: Tokenomics 101 — Eugene Cheah, Featherless AI →
Links
GitHub 
X Featherless Co-founder and CEO · Featherless
Career
Featherless began as Recursal.AI, founded by Cheah, Harrison Vanderbyl and Wesley George after work on RWKV. RWKV explores attention-free model architectures. Getting open models into production exposed another obstacle: a published model can still be difficult for an application developer to run. Featherless addresses that operational gap with a common serving API. The company also identifies Cheah as an earlier Uilicious co-founder.[1]
Ideas
Measure the workload before choosing capacity
In his AI Engineer talk, Cheah recommends separating usage by team and task before optimizing a shared AI bill. A developer waiting for an answer has different latency needs from a background batch job. He distinguishes long-running agents, which repeatedly process large contexts, from high-volume short decisions. The appropriate model and deployment follow from those workload measurements.[3]
Pay for a useful decision
Cheah also discusses classifiers that choose among options without generating a prose response, and smaller models evaluated against a defined task. His central test is useful output and the effort needed to review it, rather than token volume alone.[3]
Tools, writing & talks
Writing, research & talks
Start here
A practical talk on usage attribution, model choice and the economics of serving different AI workloads.
Start with the distinction between interactive and background workloads. Familiarity with an API bill helps; the talk explains how task requirements affect model and capacity choices.
Speaker[3]Original source
Tools
- Featherlessfeatherless.ai
An API platform that makes open models available without managing a separate deployment for each one.
Co-founder and CEO; company product[1]