Inspiration
Baho AI was inspired by the reality of Rwanda’s healthcare system, where Community Health Workers (Abajyanama b’Ubuzima) provide essential care but remain highly overextended. With over 45,000 CHWs serving more than 13.4 million people, timely and personalized health guidance is still a challenge. Rising non-communicable diseases and gaps in health literacy further emphasized the need for an always-available, accessible, and culturally grounded health support system. Baho was created to extend healthcare access using AI that speaks Kinyarwanda and aligns with Rwanda’s healthcare structure.
What it does
Baho AI is a real-time, multimodal health companion that supports users through chat, voice, and live video consultation. It allows people to interact with Baho in Kinyarwanda, ask health questions, receive guidance, and even show physical symptoms or situations through video for contextual understanding.
It supports structured long-term health journeys through:
9 Months with Baho — maternal health tracking and guidance 1000 Days with Baho — child growth and development monitoring Baho for Life — chronic disease management and continuous care
Users can now call Baho via live video, enabling real-time visual interaction where Baho can observe, respond, guide, and support decision-making in a more immersive and practical way.
How we built it
Baho AI was built as a full-stack, AI-driven health ecosystem with three integrated layers:
User Experience Layer: Mobile-first interface with chat, dashboards, health programs, and real-time video call capabilities AI Intelligence Layer: Context-aware conversational system that uses memory, health program data, and real-time interaction inputs (text, voice, and video) Health Data Layer: Structured longitudinal databases for pregnancy, child development, chronic care, and interaction history
The platform integrates real-time communication systems to support live video consultations, enabling Baho to process and respond to visual and verbal health inputs dynamically.
Challenges we ran into
Designing real-time AI video interaction that feels natural and responsive Ensuring safe and responsible interpretation of visual health inputs Structuring multimodal data (text, voice, video context) into meaningful health responses Maintaining cultural and linguistic accuracy in Kinyarwanda during real-time conversations Balancing responsiveness with medical safety constraints
One of the hardest challenges was making real-time visual interaction feel helpful without overstepping medical boundaries.
Log in or sign up for Devpost to join the conversation.