Ryan Heath interviews Peter Kant, CEO and co-founder of Enabled Intelligence, about revolutionizing AI data labeling through neurodiverse talent.

Peter shares how his company solves a critical bottleneck in AI development — high-quality labeled training data — while creating meaningful employment for neurodiverse individuals and people with disabilities.

From achieving 95% accuracy rates compared to the industry standard of 70% to developing thin AI models for edge deployment, this conversation reveals how diversity in human cognition creates more robust, efficient, and representative AI systems that benefit both national security and commercial applications.

Key Takeaways

Neurodiversity Drives AI Quality and Efficiency

Enabled Intelligence's workforce is over 50% neurodiverse or persons with disabilities, leveraging hyperfocus, pattern recognition, and attention to detail — delivering 95% accuracy in data labeling versus the 70% industry standard, Kant says, while processing data two to three times faster than typical workforces.

High-Quality Training Data Reduces AI Costs Dramatically

Better labeled data consumes less compute power. When training data contains errors, AI systems must learn workarounds, while representative, accurately labeled data creates lighter, more efficient models that can operate at the "edge" without massive infrastructure.

Brain Diversity Creates More Representative AI

Successfully mimicking human thought through AI means mimicking more than software developers from Stanford. By incorporating neurodiverse perspectives in data labeling, Enabled Intelligence's training data better represents the spectrum of human cognition, resulting in more reliable AI models.

Specialized AI Tools Are the Growth Frontier

Enabled Intelligence has expanded into model fine-tuning and development, creating purpose-built, lightweight AI tools for specific business needs, from proposal writing to electronic medical record analysis.

Professional Workforce Model Pays Off

Higher labor costs in the U.S. are offset by high retention rates, and low error rates, which delivers enough efficiency and stability to make the economics work.

Hyperspectral Imaging Unlocks Hidden Intelligence

By combining hyperspectral satellite imagery — capturing roughly 220 different light spectra — with AI analysis, previously impossible applications become feasible. From identifying lithium mines and monitoring deforestation to detecting camouflaged military assets, AI now processes what was impossible or previously very labor-intensive to identify.

Chapter Timestamps

[00:00] Introduction and company mission

[02:00] Origin story at Stanford Research Institute

[04:00] The data labeling bottleneck problem

[06:00] Israeli cyber battalions inspiration

[08:00] Economics of neurodiverse workforce

[10:00] Accuracy rates and efficiency gains

[13:00] Model fine-tuning and specialized AI

[17:00] Hyperspectral imagery explained

[22:00] Company expansion

[24:00] Recruiting and training approach

Peter Kant's computing background is grounded at Stanford Research Institute (SRI International), where Peter identified a critical gap in the AI ecosystem: the lack of access to reliable, accurately labeled training data, particularly for classified and sensitive applications.

Drawing inspiration from Israeli Defense Forces' cyber battalions that employed neurodiverse soldiers, Peter built Enabled Intelligence with a workforce that is majority neurodiverse or people with disabilities. The company has expanded beyond data labeling into AI model fine-tuning and development, creating specialized, lightweight AI tools for both defense and commercial applications. The company recently doubled in size over two months and is expanding operations from its base to St. Louis, with interest from NATO countries.

Connect with Peter Kant

https://enabledintelligence.net/

https://www.linkedin.com/in/peterkant4/https://enabledintelligence.net/our-people/