The intellectual property of HeyDonto AI Technology in one place: peer-reviewed publications, research under review, the patent estate, and engineering notes from across Axiomera, Conduit, DFT Labs. Most of it is deep research into foundational methods: quantum-inspired models and frameworks, the geometry of learning, and deep learning applied at clinical scale. Every item links to its full text.
Every paper is first-authored by Dr. Reza Nehzati, who publishes as Mohammadreza Nehzati. All are sole-authored except the wearable sensing paper, where he is first author with co-authors.
The paper behind the company's allowed U.S. patent: evolutionary optimization that tunes deep networks against large, messy real-world data.
Code-generation systems that detect and repair their own failures.
Architectures that continuously adapt as their data environment changes.
Privacy-preserving prediction where the model travels and the data stays put.
The published foundation of the Data Field Theory research program: learning as geometry on Riemannian manifolds.
Spotting pedestrians in surveillance video reliably, even with occlusion and noise.
CNNs for software engineering under collaborative and unbalanced data conditions. Sole author.
A wearable gas sensor on patterned graphene. First author, with co-authors.
Full-text research posts for work currently in journal review. Status is stated on every piece and never overstated.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Full text available as a research post.
Why systems that exchange data flawlessly can still disagree about what it means, and what a semantic layer actually has to do.
The lab's model architecture, documented as technical reports. Internal research, running in production.
The engine that classifies raw healthcare fields against shared clinical ontologies and adapts to each customer's data.
Synthesizes mappings between healthcare data standards in both directions, rather than maintaining hand-built crosswalks.
Keeps multi-site healthcare data harmonized continuously, repairing drift as sources change.
Orchestrates the lab's classification, mapping and harmonization models as one system at production scale.
Operators on a concept Hilbert space that enrich themselves as they classify. The most formal statement of meaning as geometry.
Allowed by the USPTO and published as US20260030487A1. Additional applications are pending across the portfolio's semantic classification, mapping, and harmonization methods.
HEYDONTO word mark (Reg. 8,050,256) and logo (Reg. 8,093,112); registered copyrights covering the ENN research paper and the HeyDonto API overview.
Working notes from Conduit and Red Wheelbarrow engineering: PHI encryption and tagging, audit logging, tenant isolation, and agentic end-to-end QA. First entries are publishing here shortly.