Cognitive AI Systems for High-Precision Decision Making Case Study: Medisync

First things first!

Submit your paper/talk for the Africa Deep Tech Conference 2026 (in partnership with The Omniverse Africa). Deadline December 22, 2025!

Cognitive AI Systems for High-Precision Decision Making Case Study: Medisync

At a recent Africa Deep Tech Community Meetup, the team from Medisync presented their ideas on leveraging Cognitive AI Systems for High-Precision Decision-Making in the medical space. This was facilitated by Oyekan Remilekun and Ogunlokun Victor.

Cognitive AI for African Healthcare

Remilekun’s presentation focused on MediSync, which addresses the doctor shortage and long wait times in African healthcare. She explained that Africa has a ratio of 600 patients to 1 doctor, leading to delayed diagnoses and patients often avoiding hospitals due to long wait times. The current solutions, such as telemedicine, only provide a temporary fix by utilizing already stretched medical professionals.

AI Framework for Healthcare Access

The team discussed their work on ADAM, a framework designed to improve the functionality of language models by enabling them to understand and reason with concepts rather than just making statistical predictions. Victor explained that ADAM addresses the limitations of current language models by using a graph-based RAG system, which allows for better encoding and retrieval of information across domains, with a specific application in medicine. Bashir highlighted ADAM’s goal to democratize AI by making models more accessible and customizable to different environments and languages, particularly in Africa. Remilekun emphasized that their solution, MediSync, aims to address the doctor shortage and long wait times in African healthcare by providing an AI-powered diagnosis tool that can understand and reason through symptoms in real-time, offering preliminary diagnoses and directing users to nearby hospitals for further care.

AI Telemedicine Solution for Africa

MediSync, an AI-powered telemedicine solution integrated with ADAM, which enables users to describe symptoms via chat or call to receive an instant preliminary diagnosis and recommendations for next steps. They highlighted Medicine’s competitive advantages, including its combination of ADAM and large language models, localization for African languages, and scalability for partners. The team shared their progress, including a demo-ready system, participation in the Meta Accelerator program, and partnerships with 11 Labs, Google Cloud, and Amazon. They requested investment to enhance product development, expand their user base, and secure endorsements from the Federal Ministry of Health, while aiming to empower accessible healthcare in Africa.

Adam Architecture: Flexible Language Models

Chukwuemeka led a discussion on AI frameworks, with a particular focus on the Adam architecture. Victor explained that while small language models are portable and can be fine-tuned, they have limitations in context length. The Adam architecture, however, can work with any functional language model and encodes external knowledge without requiring fine-tuning or a rigid pipeline. Victor emphasized that Adam allows for iterative reasoning and the ability to choose among options, making it more flexible than traditional statistical language models.

AI Model Benchmarking

The team discussed benchmarking approaches for their AI model, with Alexander suggesting comparisons to ChatGPT and small language models based on quality, training process, accuracy in the Nigerian context, and inference cost. Victor explained their current benchmarking efforts, including using OpenAI’s health benchmark for emergency responses, though they are still seeking qualified mathematics teachers for more comprehensive testing. The team also addressed concerns about healthcare data sensitivity, with Victor describing their ADAM framework’s ability to learn and unlearn, and team member Dr. Nasir joining to provide medical expertise and validation of the model’s recommendations.

Medical AI System Accuracy

The team discussed their medical AI system’s accuracy and revenue model. Dr.Nasir explained that while the system aims for 80-90% accuracy in pre-diagnosis, it still requires human validation in hospitals, and they use a “Human in the Loop” approach with medical experts to improve the model. Remilekun detailed their revenue model, which includes a premium subscription plan with a 48% gross profit margin, while Bashir and Victor explained they chose healthcare over mathematics due to easier expert availability and better validation possibilities, noting they are focusing on recommendations and pre-diagnosis rather than full diagnosis to minimize risks. Remilekun confirmed that they are following Nigerian data protection guidelines. The conversation ended with Chukwuemeka announcing an upcoming deep tech conference in Lagos and encouraging participants to review the MediSync product.