When the Deep Learning Indaba launched its Ideathon in 2022, the ambition was to create a pan-African space where machine learning could move beyond theory and become a practical tool for solving the continent’s most urgent problems. Three years on, the 2025 winners suggest that ambition is no longer theoretical. It is operational, entrepreneurial and increasingly global in relevance.
Announced today, the Deep Learning Indaba Ideathon 2025 winners represent a cross-section of Africa’s emerging innovation economy, founders, researchers and builders applying artificial intelligence to healthcare, agriculture, education, space technology and cultural preservation. Their work reflects a broader shift underway across the continent where AI is not seen as an abstract frontier, but as infrastructure.
A Pan-African Engine for Innovation
The Ideathon is a flagship initiative of the Deep Learning Indaba, Africa’s largest community dedicated to artificial intelligence and machine learning. From the outset, it was designed to encourage cross-border collaboration. Each participating team must include members from at least two African countries, embedding a pan-African logic into the innovation process itself.
This year’s programme followed a refined three-phase structure. Participants moved from team formation and expert mentorship during the Indaba, to structured feedback through the DLI Mentorship Programme, before submitting final videos and participating in live Q&A sessions with judges. The result, organisers said, was “Exceptional projects addressing real-world challenges in healthcare, agriculture, education, space technology and cultural preservation”.
Those challenges are not abstract. Across Africa, healthcare systems are under strain, smallholder farmers face climate volatility, satellite congestion is rising and cultural heritage risks being flattened by Western-trained AI models. The 2025 Ideathon winners engage directly with these realities.
Research Track Winners: Rethinking What AI Sees and Misses
Bias: Endangering Species and Model Performance
Winners: Taliya Weinstein, Austin Kaburia, Kevin Kibaara
Across Africa’s protected areas, camera traps generate millions of wildlife images each year. Yet up to 75 per cent are false triggers and the species most critical to conservation, “endangered, nocturnal, and rare,” are precisely the ones AI models fail to detect.
Tools like SpeciesNet perform well on common animals yet miss the majority of leopards, pangolins and other vulnerable species. These failures are hidden inside global accuracy metrics but have profound consequences for conservation policy and population monitoring.
Their solution integrates IUCN threat metadata, strengthens representations of rare species using taxonomic and few-shot fine-tuning and introduces uncertainty-aware models that flag doubtful cases for ranger review. The aim is not just better accuracy, but better decisions on the ground.
IkirereMesh — Ikirere Orbital Labs Africa
Winners: Jason Quist, Gideon Salami, Ignatius Balayo, Jessica Randall, Alph Doamekpor
As satellite traffic increases globally, orbital congestion is becoming a real risk, one that Africa is entering at a moment when space access is democratising.
IkirereMesh is a machine-learning system that helps satellites avoid collisions and coordinate safely. It monitors satellite movement, predicts close approaches and recommends small orbital adjustments. The system operates through a ground-based planner for simulations and a lightweight version that can run directly on CubeSats.
The ambition is explicitly developmental.
“Our mission is simple but ambitious: enable African universities and research labs to learn, experiment and deploy safe CubeSat systems, contributing meaningfully to global space sustainability,” the team explains unlocking local capacity for climate monitoring, connectivity and scientific research.
Applications Track Winners: AI as Public Infrastructure
DawaMom — Dawa Health
Winners: Tariro Munzwa, Khanyisile Magagula, Kudzai Mwedzi
In low-resource settings, maternal health and cervical cancer prevention often fail not because solutions are unknown, but because access is fragmented. DawaMom positions AI as connective tissue.
The platform offers personalised pregnancy support, symptom triage and health education, while using machine-learning tools for early risk identification. For cervical cancer, it supports midwives through AI-assisted visual evaluation and clinical decision support, improving screening accuracy where specialists are scarce.
By integrating with community clinics, DawaMom enables timely referrals and continuity of care. The model is human-centred by design. AI is assisting, not replacing frontline health workers.
AuthMed
Winners: Georges Byona, Abimbola Abe, Abraham Imani Bahati, Valarie Chebet
Counterfeit medicines remain a silent killer across African markets. AuthMed confronts this with an AI-powered mobile application that allows instant drug verification.
Using computer vision and multi-modal machine learning (YOLO), the app analyses packaging against a trusted database to detect fakes. Designed for Africa’s fast-growing smartphone market, it offers a universal, scalable safety tool that empowers patients, healthcare professionals and regulators alike.
Honourable Mentions: Where Africa’s Next Ideas Are Taking Shape
From culturally grounded generative AI to climate-resilience platforms, the honourable mentions reflect both breadth and depth.
AfriVerse AI is training culturally aware generative models on authentic African film data to correct persistent misrepresentation in global AI systems. Starting with Baganda culture, the project fine-tunes open-source models using expert-validated annotations of attire, rituals and language.
Africa_Voice and Utter both tackle speech recognition, but from different angles, one through multimodal ASR combining audio and visual cues, the other by translating Kenyan non-standard English and Swahili speech patterns into clear speech in real time, improving inclusion for people with speech impairments.
In education, The African STEM Resources Hub and Team AfriMeducate address structural gaps in learning, offering offline-capable platforms, AI-powered translation, virtual labs and locally relevant medical simulations focused on diseases like malaria and Lassa fever.
Climate and food security feature prominently. Nuruuu delivers community-centred landslide forecasting in Eastern Africa, while SenseAgri AI applies causal digital twins to poultry farming, replacing trial-and-error with predictive, welfare-aware decision-making. Team Ikimera focuses on crop disease detection for smallholder farmers, addressing yield losses that currently reach 40 per cent in parts of Rwanda.
Cultural preservation is treated not as nostalgia but as innovation. Ghinel (Digital Griot) transforms African oral history into interactive, voice-based digital experiences. Its MVP chatbot, centred on King Behanzin, demonstrates how AI can preserve memory while engaging younger generations.
Healthcare inclusion is central to SignCare, which enables two-way communication between clinicians and Deaf or hard-of-hearing patients using AI-powered computer vision and speech recognition localised for Rwanda Sign Language and built with privacy at its core.
A Signal Beyond the Winners
Taken together, the Ideathon projects reflect wider global trends where AI is moving closer to users, increased emphasis on trust and explainability and a growing recognition that local context matters. What distinguishes this cohort is not just technical competence, but intent.
“These projects showcase the incredible talent, creativity and commitment to solving African challenges with African solutions,” Deep Learning Indaba said.
For investors, policymakers and entrepreneurs watching Africa’s innovation economy, the signal is clear. The next wave of global AI solutions is not only emerging from the continent it is being shaped by its realities, values and ambitions.
