HeyDonto is applying knowledge graphs to healthcare, representing a seed vertical AI play with core generative AI integration.
As agentic architectures emerge as the dominant build pattern, HeyDonto is positioned to benefit from enterprise demand for autonomous workflow solutions. The timing aligns with broader market readiness for AI systems that can execute multi-step tasks without human intervention.
HeyDonto offers AI solutions for dental EHR integration, supporting systems like Dentrix and OpenDental.
Axiomera — a patent‑pending, peer‑reviewed semantic intelligence platform that automates harmonization of ambiguous EHR data using evolutionary neural networks plus generative AI to produce AI-ready, consistent, bi‑directionally synchronized records across on‑prem and cloud EHRs.
Implied use of semantic/entity-level harmonization and a unified data model that links patient, treatment, scheduling and billing concepts across systems. They position Axiomera as a semantic intelligence layer that performs classification, mapping and harmonization — functionality commonly implemented with graph representations or graph databases to model relationships and permissions.
Emerging pattern with potential to unlock new application categories.
Generative models are used to convert unstructured clinical inputs into structured artifacts: procedure codes (CDT), standardized notes, and mapping rules to FHIR/HL7. This expresses NL->structured formats and rule/code generation rather than only free-text summarization.
Emerging pattern with potential to unlock new application categories.
The system describes validation, compliance and data-quality controls (RBAC, audit logs, detectors for missing/conflicting data). While not explicitly called an LLM-based guardrail, these components likely act as secondary validation/monitoring layers ensuring outputs meet safety and regulatory constraints.
Accelerates AI deployment in compliance-heavy industries. Creates new category of AI safety tooling.
Live EHR ingestion, near-instant synchronization and ENNs that iteratively optimize imply closed-loop feedback: production usage and real-time data are used to retrain/refine models and push improvements back into the system (a continuous-learning flywheel).
Winner-take-most dynamics in categories where well-executed. Defensibility against well-funded competitors.
HeyDonto builds on LLMs, RAG, Evolutionary Neural Networks (ENNs). The technical approach emphasizes unknown.
Compound at the service/pipeline level: ENNs and generative models function as distinct microservices within an event-driven data pipeline that performs ingestion -> semantic normalization -> enrichment -> predictive serving -> generation; explicit model-to-model routing or in-model chain-of-thought orchestration is not documented.
Serial entrepreneur; co-founded CloudRetriever (intelligent VOIP platform) with exits; founded Vet Marketing Pro and Veterinary Dashboards; UC San Diego alumnus
Previously: CloudRetriever, Vet Marketing Pro, Veterinary Dashboards
Co-founded CloudRetriever with Rivers Morrell; leads GTM at Axiomera; track record in building and scaling tech businesses
Previously: CloudRetriever
Strong: founders have prior exits and demonstrated domain alignment with healthcare data interoperability, dental EHR workflows, and partnerships with data aggregators and analytics platforms.
sales led
Target: enterprise
custom
field sales
• Targets DSOs and health-system customers.
• References data aggregators, analytics platforms, and dental distributors as collaboration targets.
Real-time interoperability and data synchronization between EHR systems to enable AI-driven automation and analytics for dental practices.
Using ENNs as an explicit layer to iteratively optimize and harden predictive models before or alongside generative LLMs is uncommon in mainstream deployed stacks and suggests an emphasis on automated architecture/parameter search integrated into the inference/enrichment pipeline.
The tight integration of semantic normalization (peer-reviewed, patented) with per-field two-way sync into live EHRs is uncommon. The claim of mathematically specified, production-grade semantic intelligence emphasizes treating upstream semantic ambiguity as the root cause of downstream AI failures, which is a stronger architectural stance than many startups take.
HeyDonto operates in a competitive landscape that includes Redox, InterSystems (IRIS / Ensemble) / traditional integration engines, Mirth Connect / NextGen integration tooling.
Differentiation: HeyDonto emphasizes AI-driven semantic normalization (Axiomera), evolutionary neural nets and generative AI to automatically map/clean ambiguous data, plus a stated patent-pending platform for production-grade semantic intelligence and two-way, immediate read/write synchronization targeted initially at dental EHRs and DSOs.
Differentiation: InterSystems is a general-purpose integration/DB/processing stack; HeyDonto claims a domain-focused semantic layer (Axiomera) backed by peer‑reviewed math and patents plus ENNs and generative AI built specifically to resolve cross-site semantic ambiguity and automate downstream AI-ready data and workflows for dental practices.
Differentiation: Mirth is tooling that requires human rules and mappings; HeyDonto positions itself as an active automated harmonization platform that uses ENNs and generative models to detect missing/conflicting data, auto-map to FHIR/HL7, and provide bi-directional realtime consistency without manual rework.
They position 'semantic failure' (ambiguous, inconsistent source data) as the primary bottleneck and have engineered a dedicated platform (Axiomera) to canonicalize semantics before any ML inference — suggesting a math-first, infrastructure-led approach rather than model-first.
Integration of Evolutionary Neural Networks (ENNs) with generative models for both data harmonization and downstream automation — implying evolutionary algorithms are continuously tuning mapping / normalization models, not just predictive models.
Claimed end-to-end, two-way read/write coverage for both cloud and on‑prem EHRs with near-instant updates 'to every table and field' — this implies bespoke connectors, MLLP/HL7v2 adapters, FHIR gateway logic, and a real-time change data capture (CDC) + conflict resolution layer operating at large scale.
Axiomera is presented as a mathematically-specified substrate (peer-reviewed) plus an 11-patent portfolio — a deliberate attempt to turn semantic mapping / harmonization into defensible IP rather than open connector engineering.
SCE / DMM / DHM / EMI pipeline (unnamed in detail) suggests a layered semantic engine architecture: substrate (SCE) for representations, mapping (DMM), harmonization (DHM), and an event/messaging/inference layer (EMI) — an unusual explicit semantic pipeline nomenclature.
HeyDonto's execution will test whether knowledge graphs can deliver sustainable competitive advantage in healthcare. A successful outcome would validate the vertical AI thesis and likely trigger increased investment in similar plays. Incumbents in healthcare should monitor closely for early signs of customer adoption.
“Powered by Generative & Evolutionary AI”
“Generative AI capabilities automate tasks like CDT code generation, clinical note drafting, and personalized treatment recommendations”
“Automated Documentation - Reduce administrative overhead with AI-generated clinical notes, patient summaries, and reports”
“Evolutionary Neural Networks (ENNs) iteratively optimize model parameters”
“generative AI systems (LLMs, RAG, knowledge graphs)”
“Integration of Evolutionary Neural Networks (ENNs) with generative models to both optimize predictive models and perform data mapping/enrichment”
