TPinfinity×Syneos Health
How it works

From the patient's words to a verified record

One conversation, captured on an AI recording badge, becomes a quality-checked and signed record in near-real-time. Every step runs onshore.

AI recording badge worn by the PSR
AI recording badge
Pipeline · 8 stages
Raw audio on device
  1. 01 · Capture
  2. 02 · Secure upload
  3. 03 · De-identify
  4. 04 · Transcribe
  5. 05 · Rule QC
  6. 06 · LLM analysis
  7. 07 · Verify + sign
  8. 08 · Deliver
Stage 1 of 8

On-device capture

The PSR wears an AI recording badge with a visible recording light, so the patient knows recording is on.

Technical detail
  • Audio captured at 16 kHz, 16-bit
  • On-device buffering with automatic upload
  • Visible indicator supports consent transparency
Data state at this stage
Spoken conversation→Raw audio on device
Compliance rail · runs across every stage
Onshore throughoutDe-identified before analysisEncrypted in transitVendor not in the data chainOriginal audio short-retentionPseudonymized longitudinal storeAccess logs ≥ 3 years

Aligned to PIPL Art. 28 · GB/T 45574-2025 · NMPA GVP · Data localization

End-to-end latency budget

Target · validated in PoC
Capture → cloud ~3 s+Speech-to-text 0.5–2 s+Rule engine <0.5 s+Downlink to PSR 0.5–1 s=Near-real-time ~4–6 s