Record
Capture raw PPG, ECG, EEG, and IMU sessions across device platforms.
BECOMING THE FOUNDATION EVERY WEARABLE BUILDS ON
AI-native R&D infrastructure for wearable biosignals.
Record, assure, generate, build, trigger, run, deploy, and monitor — one workbench.
- Record
- Assure
- Generate
- Build
- Trigger
- Run
- Deploy
- Monitor
- PPG
- ECG
- EEG
- IMU
THE PHANTOM LOOP
The loop wearable teams don't have.
Wearable teams stitch together a notebook, a spreadsheet, and a Slack thread instead. Phantom brings the entire lifecycle into one workbench.
Capture real device sessions and quality-gate the signal before algorithm evaluation. SNR, motion artifacts, and placement confidence on every window.
Generate synthetic variation that recruited cohorts cannot reach, then build and iterate algorithm logic within one continuous environment.
Simulate real-world conditions and run algorithms across population variants for a coverage report — not just one average accuracy score.
Deploy compiled algorithms directly to target hardware and watch performance live to isolate hardware, firmware, or placement drift.
RAEH Phantom · Available now
The workbench wearable teams are missing.
RAEH Phantom is the one place a wearable algorithm gets recorded against, quality-checked, tested against variation you couldn’t recruit, built, deployed to a real device, and watched live.
Record & Assure
Quality-gate raw device signals before algorithm evaluation.
Generate & Build
Test variation and edge cases no recruited cohort reaches.
Trigger & Run
Validate algorithms against Phantom-generated synthetic signals.
Deploy & Monitor
Ship directly to target hardware and watch performance live.
Coverage vs Accuracy
A coverage report, not one accuracy number.
A single accuracy score hides where an algorithm fails. RAEH reports where it breaks, not just how well it scores on average.
At t=58s this record's own confidence bottoms at 0.91 — exactly where its PPG and ECG heart-rate estimates disagree by 9 bpm. Below the 0.95 threshold every biomarker relays to hold.
- BiomarkerState
- AFibtrusthold
- Heart ratetrusthold
- HRVtrusthold
- SpO₂trusthold
- Respirationtrusthold
- Sleep stagingtrusthold
- Activitytrusthold
- Blood pressuretrusthold
read headt=58s
confidence0.91
ppg / ecg83 / 92
disagreement9 bpm
Assure
Automatic artifact detection, SNR verification, and signal quality-gating.
Generate
Create physiological, motion, and fit variation no recruited cohort reaches.
Trigger & Run
Test real-world conditions across populations for a full coverage report.
Deploy
Compile and ship validated algorithm logic directly to target hardware.
Monitor
Watch live device streams to isolate hardware, placement, or model drift.
The vision
The IDE is the wedge. Interpretation is the destination.
The workbench is the first infrastructure layer. Everything that loop produces builds the interpretation layer underneath it.
Continuous attributed sessions
The IDE is the only place a continuous, attributed stream of real biosignal data gets produced.
PPG
ECG
EEG
IMU
Runs anywhere
The same backbone across every deployment target.
- Cloudserver-side
- Mobilephone & tablet
- Wearablewatch & band
- Embeddedon-device
Population coverage
Report where an algorithm breaks across variation, not just one average score.
Signal Quality Assurance
Quality-gate raw vitals and SNR before algorithm evaluation.
Cross-modal transfer
What the backbone learns from one modality carries to the next.
- PPG
- ECG
- EEG
- IMU
- 8loop steps
- 4modalities
- 1workbench
Developer Workbench
Built for wearable algorithm teams.
When a metric moves — hardware, firmware, placement, or the algorithm? RAEH Phantom gives wearable teams one workbench to record, test, and debug regressions in code.
- PPG
- ECG
- EEG
- IMU
analyze.py
pythonimport raehraeh.api_key = "raeh_sk_live_..."# 30 s of single-lead ECG at 250 Hzecg = load_recording("session_01.npy")result = raeh.signals.analyze( signal=ecg, modality="ecg", sample_rate=250,)print(result.heart_rate)Response200 OK
biomarkersapplication/json
- heart_rate62.4bpmconf 0.98
- hrv_rmssd48.2msconf 0.94
- respiration_rate14.6br/minconf 0.91
- stress_index0.27conf 0.88
- spo2_trendstableconf 0.93
Typed SDKs
Python + TypeScript, full type hints.
Versioned & stable
Semantic versioning, no surprise breakage.
RAEH Phantom
The workbench wearable teams are missing.
Record, assure, generate, build, trigger, run, deploy, and monitor — one loop instead of six disconnected tools.
BECOMING THE FOUNDATION EVERY WEARABLE BUILDS ON