Naveen Rao is CEO and cofounder of Unconventional AI, which is exploring computing architectures that use physical dynamics to perform AI computation. His earlier companies, Nervana Systems and MosaicML, were acquired by Intel and Databricks.[1][2]
Start with Neural co-evolution →
Education
Links
Unconventional AI
X LinkedIn Cofounder and Chief Executive Officer · Unconventional AI
Career
Contributions
Unconventional AI’s introduction describes an effort to expose useful physical behavior in silicon to software, rather than simulating every state change with conventional digital operations. Rao leads the company with collaborators across algorithms, software and hardware. Biology-scale energy efficiency is the stated research goal, not a delivered performance result established by the announcement.[2]
Ideas
Design the model and the machine together
In Neural co-evolution, Rao questions treating matrix operations and existing hardware interfaces as fixed constraints for future AI. He argues for changing neural-network methods and hardware primitives together. The essay’s initial dynamical-system illustration is a simulation, with images credited to Matthew Bull; it demonstrates a research direction rather than a manufactured chip’s energy performance.[6]
Treat the workforce as part of the buildout
With Pat Gelsinger, Rao coauthored a Fortune commentary on AI, employment and infrastructure. They argue that expanding construction and technical capacity matters alongside chips and capital, drawing on their operating experience. The essay is their case for adapting skills and increasing productive capacity, including the disruptions that workers face.[7]
Tools, writing & talks
Writing, research & talks
Start here
Neural co-evolutionunconv.aiAn argument for developing AI methods and computing primitives together, with a simulated dynamical-system example.
Start with the hardware/software boundary and the dynamical-system illustration. Familiarity with neural networks helps with the examples; the initial simulation explains the research idea without demonstrating a finished hardware system’s efficiency.
Author[6]In a Fortune interview, Naveen Rao and Pat Gelsinger argue AI data center buildout is bottlenecked by skilled labor rather than chips or capital, citing a roughly 350,000-worker construction shortfall and electrician wages rising 2-4x faster than average.
Catalog workOriginal source- Introducing Unconventional AIunconv.ai
The company’s explanation of its physics-based computing research goals and approach.
Cofounder and CEO; company announcement[2] A commentary connecting AI deployment with skills, construction capacity and employment.
Coauthor with Pat Gelsinger[7]