Home Research
Our client work rests on applied research into AI that is accurate, explainable and safe to rely on. The methods we develop for high-stakes clinical settings carry straight into business systems.
VitalCKM, founded by Howard, builds governed AI infrastructure that helps clinicians and health systems see risk earlier, explain it clearly and act on it. Heart disease, kidney disease and diabetes share the same roots and accelerate one another, so VitalCKM models them together on one intelligence engine instead of three disconnected tools.
The platform predicts risk, explains which factors drive it, simulates what-if changes, tracks patients over time and integrates with health record systems through standard interfaces, all under one auditable architecture.
Four questions we keep working on, because every client system depends on the answers.
Factor-level attribution and constrained what-if simulation that show which changes move an outcome and by how much, written for the person making the decision rather than the data scientist.
In client work: lead scores, churn risk and no-show predictions that come with their reasons.
A model that says 30% should be right about 30% of the time. We test calibration and report likely ranges, so decisions account for how sure the model really is.
In client work: forecasts with ranges and alerts when reality falls outside them.
Hybrid machine learning and deep learning architectures with controlled generative data augmentation, for problems where the important cases are rare.
In client work: useful models for businesses without big-company data volumes.
Audit trails, versioning, drift detection and longitudinal tracking that keep a model trustworthy months after launch.
In client work: managed AI with monthly accuracy and drift reports.
We welcome research collaborations, hospital and clinic pilots, and academic partnerships in explainable and governed AI.