Reflections from Deep Learning Indaba 2025

The Deep Learning Indaba 2025 in Kigali, Rwanda, brought together more than a thousand participants from across Africa and beyond. For our lab, the Data Science for Social Impact (DSFSI) Research Group at the University of Pretoria, it was not just a conference but a gathering of purpose. The Indaba has always been about more than algorithms, it is about people, stories, culture, and the vision of building an African AI community that speaks in our own voices.

This year, our group reflected on moments of personal growth, meaningful connections, and the commitments we are carrying forward. Here are some of our collective experiences and insights.

A Gathering Alive with Ideas

The week began with powerful keynotes that challenged us to rethink the foundations of AI. The opening keynote, “Whose Gold?”, asked us to question whose values and definitions of “safety” are embedded in AI systems. This question became a touchstone for many of us throughout the Indaba.

• Idris found himself drawn into discussions not only about building new datasets, but also about auditing what already exists. “We cannot continue building without reflection,” he noted, citing a Hausa proverb: waiwaye adon tafiya — looking back beautifies a journey. For him, the Indaba was a reminder that progress requires careful attention to the quality, ownership, and representativeness of our data.
• Thapelo, attending after two missed years, described rediscovering the thrill of fast-paced academic exchange. A highlight for him was learning about survivorship bias in data collection, presented by Girmaw Abebe. The story of World War II planes reinforced how easily unseen gaps can distort our understanding. For Thapelo, this was an eye-opening lesson in thinking critically about the invisible in our datasets.
• Takura was struck by conversations on the immense energy demands of generative AI. Against the backdrop of Africa’s energy challenges, he reflected: “Our path forward must be rooted in leaner models, low-resource systems, and innovations designed for the continent, by the continent.”

The Human Side of Indaba

Beyond the lectures and workshops, the Indaba spirit was felt in small interactions — offers to share datasets, spontaneous GitHub exchanges, or the simple but profound question: “How can we help?”

• Dinorego highlighted the strong presence of organised machine learning communities from across Africa. Having attended previous Indabas, she was inspired to build a machine learning interest group at her own university. She also connected with Henok from Ethiopia, who generously shared insights from his team’s hackathon strategy, sparking new ideas for her PhD work.
• Fiskani described the event as transformative, beginning with the keynote on deep alignment — balancing short-term AI performance with long-term societal benefit. Conversations with innovators like Lyvia Lusiji, who presented UlizaMama (a maternal health LLM in low-resource languages), showed him what community-aligned AI can look like in practice. “Impactful AI isn’t just technical; it’s purposeful and rooted in the communities it serves,” he reflected.
• Nontokozo, attending her first Indaba and first trip outside South Africa, experienced the event on multiple levels. From learning to navigate Kigali’s motos to trying local dishes like plantain and jollof rice, she gained a deep appreciation for Rwanda’s culture. A keynote by Dina Machuve on poultry disease diagnostics stood out as an example of research directly addressing livelihoods. For Nontokozo, presenting her poster — and winning the poster prize — was a highlight that affirmed the value of her work.

For Ronald, the highlight was presenting his poster on the Smart Aquaculture Assistance System, designed to support smallholder fish farmers. His reflection captured the heart of the Indaba: “AI for human well-being is not a distant dream — it’s happening now, in rural communities.”

Our Contributions

Our lab’s presence at DLI 2025 was both visible and impactful. Members presented posters spanning topics from reinforcement learning to NLP and agriculture. Others co-organised the workshop Centering Data in African AI, which gathered voices from across the continent to reflect on ethics, governance, and benefit-sharing.

These contributions mattered because they were not just about presenting research — they were about shaping the larger conversations that will guide African AI. As Thapelo put it, co-organising the workshop meant “thinking beyond data from multiple perspectives,” and collaborating with passionate colleagues tackling African challenges head-on.

Looking Back, Looking Forward

If one theme tied our reflections together, it was the importance of community.

• We learned that building AI for Africa means addressing not just technical hurdles but questions of energy, governance, equity, and cultural representation.
• We reconnected with collaborators across countries, from Zimbabwe to Ethiopia, Namibia to Nigeria, reaffirming that our work gains strength through collective effort.
• We left with commitments: to mentor emerging researchers, to foster university interest groups, to pursue collaborations across borders, and to ensure our tools are meaningful to the people they are designed for.

As Fiskani wrote, “The Indaba has reshaped how I see my own work: not just as isolated research, but as part of an end-to-end pipeline that must lead to real-world impact.”

And as Ronald reminded us, the most transformative innovations in AI do not just process data — they see people.

Carrying the Spirit Home

From cultural dinners alive with music and spice, to poster sessions buzzing with ideas, to late-night conversations about the future of African languages and datasets, Indaba 2025 was a reminder of what happens when Africa gathers to learn and dream together.

We return to our labs and classrooms not just with notebooks full of ideas, but with renewed conviction:

• To build responsibly.
• To reflect before racing ahead.
• To make AI in Africa not only a scientific pursuit, but a human one.

The work continues, but our cups are refilled.