About EcoInference.ai
EcoInference.ai is an independent research and engineering effort focused on making on-device AI practical for everyday use — reducing AI's environmental footprint without sacrificing capability. The work is led by Mark J. Divitt and operated through Many Rivers Iowa, LLC.
Mark J. Divitt
Principal Engineer with roughly twenty years designing and building large-scale web, mobile, and cloud systems. That career was largely spent making cloud infrastructure faster, more reliable, and more capable of handling massive request volumes.
The current work is, in some ways, the opposite question: for how many of those requests was a data center actually necessary? The answer — for a large and growing share of everyday AI use — is that it isn't. On-device models have reached a level of practical capability that makes the default of routing every inference through a server worth questioning.
On honesty about AI use
Both white papers — The Case for Greener AI and AI Data Center Overbuild — were written with Claude (Anthropic) as a co-author. That's disclosed openly in each paper, because it's the honest thing to do.
Using AI assistance to research and write papers about AI is not a contradiction — it's an opportunity to demonstrate the point. The on-device equivalent of this work is one of the things we're actively building toward.
Principles
- Publish defensible numbers, not alarming ones. A floor that's climbing is more persuasive than an inflated figure that invites skepticism.
- Label derived estimates clearly. Primary sources and methodology are documented in the appendices of every published piece.
- Build in the open. This site is static, fast, and low-footprint — which is the point.
Contact
For collaboration, research inquiries, or press: info@ecoinference.ai