dr. Johan Lestari, Sp.PD-KGH

ZavoBot uses InstructGPT alignment architecture to provide safe, accurate medical responses that comply with regulatory standards.

Role-based System Instructions

Responses adapt automatically for patients with empathetic plain language and for doctors with academic clinical language.

Safety Tier Resolver

Response caution is adjusted according to medical concern classification and AI response variation limits.

Audit Log and Refinement

User feedback and clinical evaluations by doctors are analyzed to refine instructions regularly.

Role-aware Context Inclusion

Relevant Zavora medical features are recommended contextually according to the user concern.

Hallucination Guard

False diagnosis risk is reduced through clear limitations and medical fallback responses.

Data Protection and Health Regulation Compliance

Patient medical data privacy is protected according to formal healthcare operating standards.

Response Variation Limits

Response variation is limited for serious concerns so clinical information remains accurate and reliable.

Safety and Completeness Balance

The system balances patient safety protections with the completeness of AI information.