AI & Emerging Technology Lead @ HISP Rwanda
Get in touch nsanzimfurakomakech@gmail.com
Check out My Resume
I'm the AI & Emerging Technology Lead at HISP Rwanda, where I integrate machine learning, LLMs, and predictive modeling into DHIS2-powered digital health platforms used by the Ministry of Health, the Rwanda Biomedical Centre, and community health workers.
I hold an MS in Engineering Artificial Intelligence from Carnegie Mellon University. Before HISP, I advised the Government of Rwanda on national data and AI policy as a Data Science Fellow at Cenfri, and led federated-learning research on neonatal health at CMU's CyLab Africa.
I'm comfortable across the full stack — from privacy-preserving Edge LLMs and data pipelines to offline-first mobile tooling for community health workers and cloud deployment.
When I'm not coding, I explore new technologies, read behavioral psychology, and follow a bit of F1.
Lead AI/ML and LLM integrations across DHIS2-based digital health products — predictive models for climate-sensitive diseases, RAG assistants for health workers, and offline-first CHW tooling — with the Ministry of Health and Rwanda Biomedical Centre.
Advised on amendments to Rwanda's national data & AI policy and rolled out cross-departmental data standardization, improving national reporting accuracy by 35%.
Architected and led end-to-end delivery of SkillSeed, an AI-powered learning platform — microservices on NestJS, MongoDB & React with OpenAI-driven recommendations, deployed on AWS.
Led HumekaFL, a multi-hospital federated-learning project for neonatal asphyxia detection — built a Spark audio pipeline and a ResNet-50 classifier (96% accuracy) served on AWS SageMaker.
Shipped payment, deeplink-marketing, and recurring-installment features for the KreditPlus mobile app, lifting on-time payments by 15% and engagement by 13%.
A privacy-preserving RAG framework (Mistral + Chroma DB) for local, on-device interaction with Electronic Health Records — built for low-connectivity clinical settings, keeping all PHI on-device.
HumekaFL — a Federated SVM pipeline detecting birth asphyxia from infant-cry audio, enabling multi-hospital collaboration without ever moving patient data. Published on arXiv.
An agentic natural-language interface over IoT/SCADA sensors so non-technical operators can query and control farm equipment in plain language — hitting 90–100% task-completion accuracy.
An AI-powered platform helping learners aged 6–18 discover their passions and build future-ready skills through personalized learning paths.
A deep-learning model generating Kinyarwanda folk stories to preserve linguistic heritage, using BERT and KinyaBERT embeddings to produce coherent text.