We are really excited to share the winners of the 4th edition of the Deep Learning Indaba’s Ideathon. The competition this year has been strong, with a record 54 submissions. All teams can be incredibly proud of their efforts.

Below, we list and congratulate the winners for the Research and Applications tracks, along with the Honourable Mentions.

Research Track

Bias: Endangering species and Model Performance – WINNER

Taliya Weinstein , Austin Kaburia, Kevin Kibaara

Camera traps across Africa generate millions of images, but up to 75% are false triggers, and the species most critical to conservation, endangered, nocturnal, and rare, are the ones current 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 mistakes are buried in global accuracy metrics but profoundly distort population monitoring.

Our project addresses this by integrating IUCN threat metadata, strengthening representations of rare species through taxonomic and few-shot fine-tuning, and building uncertainty-aware models that flag doubtful cases for ranger review, improving reliability and field impact.

Primary contact details: 

Presentation

Ikirere Orbital Labs Africa – WINNER

Jason Quist. Gideon Salami, Ignatius Balayo, Jessica Randall, Alph Doamekpor

IkirereMesh – ML-Driven Satellite Safety & Coordination for Africa:

IkirereMesh is a system that helps satellites avoid collisions and work together safely. It watches how satellites move, predicts possible close calls, and suggests small adjustments to keep them at a safe distance. The system has two parts: a ground-based planner that runs detailed simulations, and a lighter version that can run directly on CubeSats. Our goal is to make space operations safer and more accessible for African universities and research labs. With IkirereMesh, African teams can learn, experiment, and even launch their own small satellites for climate monitoring, connectivity, and scientific research.

Website: https://ikirere.com/ 

Primary contact details:

Presentation

AfriVerse AI – Honourable Mention

Namutebi Esther, Nakabuuka Regina Desire, Nakazibwe Jovia, Moses Tijesuni Samuel

AfriVerse AI is building a culturally-aware generative AI model trained on authentic African film data to improve the representation of African traditions, attire, language, and ceremonies. Current AI systems often misrepresent African culture due to Western-centric datasets, producing inaccurate visuals for items like gomesi, kanzu, or Maasai attire. Our pipeline extracts frames and subtitles from open-licensed African films, transcribes with Whisper, and uses expert validation to annotate attire, rituals, and proverbs. We then fine-tune open-source models using LoRA adapters. Beginning with Baganda culture, AfriVerse will scale across Africa, preserving heritage, enabling creativity, and ensuring AI sees Africa as it truly is.

Primary contact details:

Presentation
Video Presentation

Africa_Voice – Honourable Mention

Nabil Badri, Ghayth Bouzayani, Praise Amonye, Mahmoud Khaled Abdelkader, Ayodele Awokoya

Our idea is to build a Robust ASR system by adopting a multi-modal approach. In addition to audio, the system will leverage visual cues such as lip movements and facial expressions. By fusing audio and visual information, the model can recognise speech more accurately and reliably, even in challenging acoustic conditions.

The motivation is to overcome the limitations of traditional ASR, especially in noisy environments or for underrepresented languages and accents. This work will enhance digital inclusion, improve accessibility for people with disabilities, and enable voice-driven applications in education, healthcare, and communication. Its impact will be to empower communities globally by making speech technology more fair, inclusive, and practical in real-world settings.

Primary contact details:

Presentation
Video Presentation

Applications Track

Dawa Health – WINNER

Tariro Munzwa, Khanyisile Magagula, Kudzai Mwedzi

DawaMom is an AI-driven digital health platform to improve maternal health and cervical cancer prevention:

DawaMom is an AI-driven digital health platform designed to enhance maternal health and cervical cancer prevention in low-resource settings. It provides women with personalised pregnancy support, symptom triage, and health education while enabling early risk identification using machine-learning tools. For cervical cancer, DawaMom supports midwives through AI-assisted visual evaluation and clinical decision support, improving screening accuracy where specialist access is limited. The platform integrates seamlessly with community clinics, ensuring timely referrals, follow-up care, and continuity of services. By combining mobile access, AI models, and human-centred design, DawaMom helps reduce delays, improve outcomes, and strengthen primary healthcare delivery at scale.

Website: https://dawa-health.com

Primary contact details:

Pitch Deck

Team AuthMed – WINNER

Georges Byona, Abimbola Abe, Abraham Imani Bahati, Valarie Chebet

Counterfeit medicines claim thousands of lives across Africa annually. AuthMed combats this crisis with an AI-powered Mobile Application that allows users to verify drug authenticity instantly. Leveraging Computer Vision and Multi-Modal Machine Learning (YOLO), the app analyzes packaging and visual features against a trusted database to detect fakes with high accuracy.

Designed for Africa’s growing smartphone market, AuthMed outperforms limited competitors by offering a universal, scalable safety tool. By empowering patients and healthcare professionals to identify substandard drugs, AuthMed aims to save lives, restore trust in health systems, and provide crucial data to regulators for better policy-making.

Primary contact details:

Pitch Deck

The African STEM Resources Hub – Honourable Mention

Muhigiri Ashuza Albin, Nibigira Danone, ‎Kibogora Nsoro, ‎Nematou Labore

The STEM Resources Hub is an educational platform designed by Africans to address the continent’s STEM education crisis. It offers accessible, localised content and AI-driven solutions. Key features include a mobile-first, offline-capable architecture; a virtual lab for simulations; and AI-powered translation into African languages. The platform provides personalised learning paths and bridges infrastructure gaps, making world-class science accessible to every African child. With immense potential to transform education, it addresses high dropout rates, lack of qualified teachers, and gender inequalities, unlocking Africa’s human potential.

Primary contact details:

Pitch Deck
Pitch Video

Team AfriMeducate – Honourable Mention

Similoluwa Okunowo, Comfort Akanni, Khalifa Babacar, Nathaniel Mugenyi

AI-Powered Simulation-Based Learning for Medical Students in Africa: 

Medical students in Africa are graduating without enough hands-on practice, and most schools cannot afford simulation centres. Textbooks often focus on Western diseases instead of local conditions like malaria, typhoid, or Lassa fever, creating a dangerous gap between training and real-world needs. Our mobile app closes this gap by using generative AI to create locally relevant simulated patient cases and provide personalised feedback. Students can safely practice history taking, patient-centred communication, and clinical reasoning anytime, building competence before meeting real patients. Our mission is to deliver low-cost simulation-based learning to every medical student in Africa.

Primary contact details:

Pitch Deck
Pitch Video

Nuruuu – Honourable Mention

Choukouriyah Arinloye, Kadidja Janny Pombot Fall, Alekhya Kasireddy

AI-Based LandSlide Detection in Eastern Africa: 

Landslides are an escalating hazard across Eastern Africa due to changing rainfall patterns, rapid land-use change, and limited monitoring capacity. Recent extreme events in 2023–2024 highlight the urgent need for early warnings that are accurate, trusted, and actionable. Nuruuu is a community-centred landslide risk intelligence platform that delivers short-term, localised forecasts through intuitive mobile alerts and dynamic hazard maps. Machine-learning models integrate rainfall, soil, terrain, and historical landslide data, supported by real-time environmental sensors in high-risk zones. Community reporting, local-language alerts, and public education ensure warnings translate into timely action. By combining predictive science with human-centred design, Nuruuu shifts landslide management from reactive response to proactive community resilience.

Primary contact details:

Pitch Deck
Pitch Video

Utter – Honourable Mention

Ephraim Mwereza, Amaziah Marvel, Emmanuel Eddy

Inclusive ASR for Kenyan English and Swahili for People with Non-standard Speech:

Utter is a speech recognition startup and social enterprise that has developed a proprietary automatic speech recognition (ASR) technology that translates Kenyan non-standard English and Swahili speech patterns into clear speech in real-time. This technology enables children and adults with speech impairments, accented Kenyan English or Swahili, or ageing voices to access mainstream voice-activated technologies and devices.

Primary contact details:

Pitch Deck
Pitch Video

Team SignCare – Honourable Mention

Nasiru Iliya and Moise Iradukunda

In Rwanda—and across much of Africa—Deaf and hard-of-hearing patients face severe communication barriers in healthcare settings. As in the case of Nancy, a 26-year-old expectant mother, many rely on paper notes or family members as informal interpreters, leading to misdiagnosis, delayed treatment, loss of privacy, and unequal access to essential care. With over 66,000 Rwandans and more than one million people globally at risk, this gap is both systemic and urgent. SignCare bridges this barrier by providing true two-way communication through AI-powered computer vision and speech recognition, enabling safe, accurate, and dignified interactions between clinicians and hearing-impaired patients—localised to Rwanda Sign Language and built with privacy by design.

Primary contact details:

Pitch Deck
Pitch Video

Ghinel – Honourable Mention

Precious Adekunle Oreoluwa, Yusupha Ceesay, Ghilth Gbaguidi, Lorraine Chepkemoi, Charmaine A. Yebadokpo , Fenu Aguidi, Ariane AGBOTON, Luc AKAKPO, Dylane LOKOSSOU SOTON

Digital Griot is an interactive platform that preserves African oral heritage by transforming traditional storytelling into a conversational digital experience. Using speech recognition, text-to-speech, it captures the rhythms, idioms, and cultural depth of local languages. Community storytellers and linguists co-create and validate content, ensuring authenticity and shared ownership. Unlike static archives, Digital Griot offers living, voice-based storytelling that reconnects communities, educates younger generations, and strengthens cultural identity. The MVP features a chatbot that answers questions about King Behanzin and demonstrates how African history can be brought to life through AI-powered, culturally rooted interaction.
Website: https://ghinel-ghinel-ghinel.vercel.app/

Primary contact details:

Pitch Deck
Pitch Video

SenseAgri AI – Honourable Mention

Dylan Geldenhuys, Cornelius (Masese) Maroa, Ryan Nel, Taliya Weinstein, Heinrich van Deventer

Correlation to Causation – Optimising Poultry Production with Digital Twins:

Poultry farming is vital for African food security but faces rising costs and thin margins. This project seeks to improve farm management by replacing trial-and-error optimisation with a Causal Digital Twin. Unlike standard dashboards that merely display data trends, our system combines veterinary knowledge, simulations, and Causal Machine Learning to predict the specific outcomes of interventions. We provide farmers with safe, actionable recommendations accompanied by transparent uncertainty gauges and welfare guardrails. Our system can learn from small-scale trials if there is insufficient historical data for recommendations. Our systems can empower farmers to improve feed efficiency and reduce mortality, ensuring sustainable, profitable production while prioritising animal welfare.

Primary contact details:

Pitch Deck
Pitch Video

Team Ikimera- Honourable Mention

Eunice Adewusi, Chambeline Nkah, Samuel Babalola, Audry Chivanga, Allen Gahigi Abarisa

Ikimera – AI-Powered Multimodal Crop Disease Detection and Precision Intervention System for Smallholder Farmers

In Rwanda, where 70% depend on agriculture yet 987,000 households face food insecurity, 40% of crop yields are lost to preventable diseases. Smallholder farmers lack specialised diagnostic tools for timely intervention. Ikimera addresses this critical gap with an offline-first, multi-modal AI platform that analyses leaf images, soil data, weather patterns, and farmer input in local languages. Acting as “AI fertilizer” for crop health, it delivers precise disease detection and actionable recommendations, reducing losses, pesticide misuse, and climate vulnerability. By increasing household incomes $200-500 annually and integrating with existing platforms, Ikimera’s scalable, human-verified system strengthens food security and resilience across East Africa.

Primary contact details:

Pitch Deck
Pitch Video