Puzzle Gen Medical Voice AI Documentation

Puzzle AI is a Seoul-based medical speech recognition company with an additional office in New Jersey, USA, led by CEO Kim Yongsik. It reduces the time doctors and nurses spend typing records in busy hospital settings such as outpatient clinics, wards, emergency rooms and operating theatres.

Puzzle Gen for physicians and for nurses captures speech and conversations with a character error rate below 1%, structures them into SOAP notes, nursing records and discharge summaries, and uses an EMR agent that opens forms, fills fields and saves them. It runs on standard PCs without GPUs and in closed hospital networks, and supports personal templates, abbreviations and voice tuning. Puzzle SDK lets partners embed the engine, and Puzzle Care supports care-team communication.

• Medical speech recognition: dictation and conversation capture with under 1% character error rate.
• Generative notes: turns conversations into SOAP notes, nursing records and discharge summaries.
• EMR agent: opens forms, fills content and saves records directly in the hospital EMR.
• Personalisation: custom call words, templates, abbreviations, formatting and personal voice models.
• Any device: works on PCs, smartphones, tablets and AR glasses with wireless microphones.
• Puzzle SDK: lets EMR and device vendors embed the speech engine.

Vietnamese tertiary hospitals could pilot voice-driven nursing records and discharge summaries in closed networks. HIS vendors could license the SDK to add medical speech input and an EMR agent to their products. Imaging and pathology departments could adopt dictation for faster reporting, as Seoul St. Mary's Hospital did.

Clinically reliable recognition with a character error rate below 1%, according to the website. Fills EMR forms automatically without costly integration projects or GPU servers. Works in closed networks, meeting strict hospital security requirements. Customer satisfaction exceeds 90%, with sustained use at more than 99% of deployed sites.

Large hospitals in Korea, Japan and Southeast Asia are moving from simple dictation to generative AI that writes and enters records, while keeping data inside closed networks. Vietnamese tertiary hospitals face heavy inpatient and nursing documentation under electronic medical record rules, creating demand for on-premise voice AI that works with existing HIS.

Offices in Seoul, Korea and Phillipsburg, New Jersey, USA. Used by more than 75% of Korean tertiary hospitals, with 750 physician and 2,250 nurse licences. Signed a smart hospital agreement with Bumin Hospital Group (June 2025), joined a Korean ARPA-H project (2025) and works with Mayo Clinic on generative medical AI.