StarDeep™ Agentic AI Platform
A generic multi-agent platform for clinical trials — secure, intelligent and scalable, from participant monitoring to nightly per-patient health summaries, configured by each study's protocol.
In development
- Indication
- Decentralized clinical trials (DCTs)
- Data source
- Fleet of connected devices (CGMs, smartwatches, phones) and apps
- Partnership availability
- Open to sponsors, CROs and clinical sites for pilot studies
The platform
The StarDeep™ Agentic AI Platform is a generic platform for clinical trials: it orchestrates decentralized studies (DCTs) with a multi-agent architecture — secure, intelligent, and scalable — and adapts to any study design, with visits, collection windows, sensors and questionnaires configured by each clinical protocol. Six cooperating agents cover the trial lifecycle end to end, delivering intelligent automation 24/7 while investigators keep the decisions:
Monitor participants
Continuous multi-sensor data from smartwatch and companion app, encrypted on upload.
Validate sensor files
A validation agent checks every incoming file against per-sensor rules — pass or fail, nothing slips through.
Extract structured features
A feature-extraction agent computes ECG quality, heart-rate variability and other clinical features.
Local LLM clinical interpretation
A medical LLM agent, hosted on-premises or in a private cloud, delivers clinical interpretation, risk assessment and insights.
Surface findings & alerts
Study portal and Grafana dashboards for study teams, with email alerts on risk escalation.
Nightly health summary
A summary-generation agent runs every night, for every participant, over all validated data.
Validated data lives in an encrypted secure data lake, with audit logs, access control and encryption throughout — compliant with HIPAA and GDPR. And no patient data ever leaves the host: clinical interpretation stays local by design.
Surface findings & alerts
- Study portal with cohort overview: participants, at-risk cases, alerts, data quality
- Grafana dashboards for operational oversight and risk distribution
- Email alerts on risk escalation — elevated heart rate, irregular rhythm, low heart-rate variability
Nightly longitudinal health summary
- Trends and metrics per participant
- Risk evolution over the course of the study
- Clinical insights and data-quality overview
- Available next morning for clinicians and study teams
Proven operational foundation
The platform is generic: it executes whatever the clinical protocol defines. The study flow — the number of visits, the length of the free-living collection window, the sensors and questionnaires — is configured per protocol, not hard-coded:
- Enrollment visit(s) — informed consent, inclusion/exclusion screening, device provisioning, eCRF entry, participant training and baseline measurements.
- Free-living data collection (N days, defined by the protocol) — participants live normally while wearing their devices, with daily check-ins on device wear and data compliance, protocol-defined questionnaires, and green/yellow/red data confirmation before sensor and questionnaire upload.
- Follow-up and closing visit(s) (N visits, defined by the protocol) — intermediate check-points as required, additional biometrics, device return, and biological testing for disease labelling.
This operational knowhow was built running NeuTigers' own wearable-based studies — including the SweetDeep™ prospective study — and generalizes to any protocol:
- International hospital-clinic networks across 4 continents, with partnerships with major KOLs and principal investigators.
- HIPAA/GDPR-compliant cloud infrastructure for security and privacy, with real-time data monitoring tools for quality and compliance.
- Large cohorts of more than 2,000 participants.
- Management of a fleet of devices — CGMs, smartwatches, phones — and their apps.
- Coordination of clinical and engineering staff: MDs, nurses, DevOps and data science.
Current status
The StarDeep™ Agentic AI Platform is in development, built directly on this operational knowhow and on NeuTigers' Princeton-origin research in subject-efficient study design. This status describes the program's development stage; it does not imply regulatory clearance, approval or diagnostic status in any jurisdiction.
Patient privacy
No patient data leaves the host.
Faster insights
Intelligent automation, 24/7.
Higher quality
Every file validated. Better data, better decisions.
Lower effort
Scalable DCTs at lower cost.
Faster insights. Higher quality. Lower effort.
Intelligent automation 24/7 for scalable decentralized clinical trials.
Collaboration need
We are looking for study sponsors, CROs and clinical sites interested in piloting secure, subject-efficient, decentralized study designs. Discuss this program with us.
Evidence
- SCouT — time- and subject-efficient clinical trials with the statistical power of resource-intensive designs (Princeton-origin research)
- SweetDeep decentralized clinical trial with AP-HP — real-world data collected in real-world conditions (arXiv publication, 2026)
Program status reflects the stage stated above; it does not imply regulatory clearance, diagnostic availability, or commercial deployment unless explicitly stated.