기술 블로그

Keep the platform open, regulate only the therapy module: Jarame’s FDA strategy is separation

Jarame will not medical-device the whole platform. It isolates functions that directly affect treatment outcomes, and layers the AI Digital Care Log into source records, RWE, and a treatment engine.

Author: 드림에이아이랩 기술팀

[Technology · 2026-08-17]

Layer the AI Digital Care Log into source records, RWE, and a treatment engine. After authorization, expand by indication.

The core of Jarame’s FDA preparation is not “make the whole platform a device.” It is independence of the treatment function. Matching, records, and education stay. Only functions that directly affect patient outcomes become separate device software.

Jarame FDA separation: an open care platform and a sealed therapeutic module divided by a clear boundary
The platform stays open. The therapy module is bounded. Conceptual illustration.

Jarame is not a single digital-therapy app. It holds treatment records, guardian observation, therapist ratings, center operations, home-therapy matching, counseling, education, person-level analysis, treatment recommendations, and long-term collection. Filing all of that as one FDA product would let a small admin change touch device change evaluation, and would explode the user types and environments that trials and usability studies must cover.

Why FDA talks about multiple-function products

When device and non-device functions live in one product, FDA may treat it as a Multiple Function Device Product. Even a non-device function can fall into safety and effectiveness review if it affects the device function’s input data, code, memory, performance, or cybersecurity. FDA evaluates data dependence, shared resources, and risk propagation, and may expect modularization. (U.S. Food and Drug Administration)

Six layers

Six Jarame FDA layers from open platform to sealed therapeutic module and quality and safety
The fifth layer is the first authorization candidate. Conceptual diagram.
  1. Jarame Platform: membership, center search, home-therapy matching, schedule, payment, counseling, communication, general education. Non-device services that do not diagnose or choose a person’s treatment.
  2. AI Digital Care Log Core: source data from guardians, therapists, centers, and hospitals, plus identifiers, consent, access, origin, time, and change history. It proves when, by whom, and on which system version a record was made, rather than judging treatment effect by itself.
  3. AI Digital Care Log Evidence: turns source data into researchable RWD and then into RWE that can answer a regulatory question. Inclusion, baseline, exposure, concomitant therapy, missingness, follow-up, comparators, and endpoints live here.
  4. Jarame Clinical: shows clinicians change, risk signals, missing data, possible interventions, and grounds. The algorithm must show which information and criteria it used, and how uncertain it is—not a bare conclusion.
  5. Jarame Therapeutic: the direct FDA subject. Content choice, order, difficulty, frequency, and next intervention that engage outcomes. The first file should freeze one age, diagnosis, goal, and endpoint—not all developmental disability.
  6. Jarame Quality & Safety: complaints, malfunctions, adverse events, CAPA, model drift, cybersecurity, and post-market safety and performance. Real-time collection can serve pre-market trials and post-market surveillance.

Separation must be technical, not cosmetic

Renaming menus is not enough. The filed module needs clear boundaries in source, database, APIs, permissions, deployment, configuration management, algorithm version, and audit logs. A platform fault must not change therapeutic calculation or output. If the two sides share data, the share range and error controls belong in risk-management documents.

Model updates change too. The first authorized product would freeze the approved model, preprocessing, and protocol as the regulatory baseline. New data would be trained in a separate research environment, then released only after independent validation and risk review. FDA’s Predetermined Change Control Plan (PCCP) lets a sponsor pre-define what performance to improve, with which data, and how to verify impact—important for a system that learns from use. (U.S. Food and Drug Administration)

Software documents become a traceability chain

Requirements, design, risk controls, code, verification, validation, and clinical evidence must connect. FDA premarket software documentation typically covers description, risk management, requirements, architecture, testing, version history, and unresolved anomalies. (U.S. Food and Drug Administration)

The gain is clear. The platform can still ship quickly. The AI Digital Care Log stays a shared RWD/RWE base across conditions. The authorized module stays strict for that indication, then the family can grow to communication, behavior, daily living, and sleep. The competitive edge is not one therapy app. It is connecting state before treatment, the intervention, everyday response, and long-term outcome—and feeding that back into improvement and post-market watch. See also the FDA entry preparation notice.

Status: This is a preparation architecture, not FDA clearance or approval. Company-stated data volumes and improvement ranges are not regulatory-grade evidence until protocol, independent analysis, and review convert them.

Related tags

  • #자람이
  • #FDA
  • #Multiple Function Device
  • #의료기기 소프트웨어
  • #SaMD
  • #모듈화
  • #PCCP
  • #QMSR
  • #AI 디지털케어로그
  • #RWE
  • #디지털치료제
  • #기능 분리

Related articles

보도자료

발달장애 플랫폼 ‘자람이’, FDA 진입 준비 본격화…10만 명 RWD로 디지털치료제 도전

자람이가 플랫폼 전체가 아니라 단일 치료 모듈을 분리해 미국 FDA 의료기기 허가를 준비합니다. 약 10만 명 RWD는 출발점이며, FDA가 인정한 임상 결과는 아닙니다.

Read more →
기술 블로그

노아AI AI 커스텀이란? 개인 투자자가 TradingView·Pine 전략을 코딩 없이 검증하고 운용하는 기술

노아AI AI 커스텀은 책·Pine Script·TradingView 전략·영상을 실행 규칙으로 바꾸고, 시장국면이 맞지 않으면 멈추며 가드레일로 위험을 제한합니다. 수익 보장이 아니라 개인 투자자의 전략 운영 계층입니다.

Read more →
보도자료

발달장애 플랫폼 ‘자람이’ 이용자 35만 돌파…홈티 시장까지 서비스 확대

자람이가 누적 이용자 35만 명을 돌파한 가운데, 프리랜서 치료사와 가정을 직접 연결하는 홈티 매칭을 확대합니다. 홈티 치료기록도 AI 디지털케어로그에 통합합니다.

Read more →