Mhina Lukurunge

I'm Mhina Lukurunge, a computer engineer and data scientist in Tanzania. I'm in the thesis year of an MSc in Data Science and Artificial Intelligence at NM-AIST, on the government's Samia Scholarship (DS/AI+). I teach computer engineering as a Tutorial Assistant at Mbeya University of Science and Technology, and I'm CTO and Director of Technology & Digital Health at Omnexus Pharma Solutions Hub, a pharmacist-led firm registered in Tanzania.

I led the engineering on Tunza Nasi, a pharmacy management platform running in 25+ pharmacies across Tanzania, each on its own isolated, encrypted database. The pharmacists I work with own the clinical and regulatory rules. Before that I was a district data officer on Afya Hatua, a CDC-funded HIV programme, leading 21 officers across TB/HIV, CECAP, GBV and PMTCT facilities, automating the DHIS2 and DATIM reporting and the quality checks behind 95%+ district viral-load coverage.

My thesis is CerviStereo, AI-automated stereology for grading cervical neoplasia on H&E biopsies. Cervical cancer is the leading cancer in Tanzania, and CIN grading between pathologists agrees at κ = 0.36. CerviStereo measures how abnormal a biopsy is as a continuous 3-D quantity with an error bar instead of a grade label, and validates those measurements against manual stereology rather than correlation alone. It runs on open whole-slide data from the Glasgow iCAIRD archive.

I also lead Mama Salama, a Swahili, offline phone companion that helps a new mother check her own and her baby's danger signs in the weeks after birth. It placed first at Pitch II of the AI4Health Local AI Buildathon and goes to the Grand Finale on 12 November 2026.

I came up through hands-on engineering before specialising in DSAI, so I don't hand a model to someone else to deploy. The common failure in Tanzanian tech is fragmentation: a separate app here, a web tool there, a USSD code somewhere else, none of them talking to each other, all abandoned within a year of the funding ending. I build integrated, offline-tolerant systems with a named owner for after I leave.

Health data is handled under written governance: quality checks before analysis, nothing patient-level outside the system it belongs to, clinical logic owned by clinicians.