Blog parent
November 2, 2026
Turning Hope into Action: The AI Research and Innovations Shaping Africa | The Encoder, Issue 19
Every great change begins with a simple idea
Across Africa, researchers, innovators and technology builders are exploring how emerging technologies can address real-world challenges. From developing AI models for African languages to improving agricultural services and strengthening digital infrastructure, their work is helping move research and innovation forward.
But developing new technology is only part of the process. Testing ideas, evaluating results and understanding what needs improvement are equally important.
In this edition of The Encoder, we explore innovations being put to the test, research shaping conversations around AI in Africa, developments in Nigeria’s digital ecosystem, and opportunities for those looking to learn, research and build.
01. TEACHING: From Learning to Building
Cohort 10 Demo Day: Three Projects Turning African Challenges into AI Research
On Saturday, 3 October 2026, the TRI AI Cohort 10 Demo Day brought together participants who had spent the programme developing and working on their projects.
Nine finalist teams presented their work, offering a glimpse into the ideas, skills and practical applications developed throughout the cohort.
The presentations highlighted how participants are applying their knowledge to real-world challenges through AI research and innovation.
The top three teams stood out for their projects and the work behind them.
Read more: Meet the Three Teams Turning African Challenges into AI Research at TRI AI Cohort 10 Demo Day
Ten Cohorts. One Question: What Difference Have We Made?
AI Saturdays Lagos started in 2018. Ten cohorts later, now as TRI AI, the organisation is taking a closer look at the impact of its work.
The question is simple: what difference has the community made in the lives and careers of the people who have been part of it?
TRI AI is inviting former cohort participants and other community members to take part in a short survey. The survey is almost entirely multiple choice and takes about five minutes to complete.
Importantly, the organisation also wants to hear from people who feel that little has changed for them. Understanding those experiences is part of measuring the community’s actual impact.
Take the survey: Share Your Experience
Help Build an AI Dataset: Share Your Kitchen Photo
Research often depends on something as simple as having enough quality data.
Israel Isiaka, a TRI AI community member and MSc student at the University of Salford, is collecting kitchen photographs for a deep learning research project.
Using a pre-trained MobileNetV2 model, he aims to develop a system that can classify kitchen images as either tidy or untidy.
The project has a target of 1,000 photographs, and community contributions can help bring that goal closer.
Anyone interested can support the research by submitting a photograph of their kitchen. No sign-up is required, and location data is automatically removed from submitted photographs.
Contribute: Kitchen Photo Collection
02. RESEARCH: The Questions Shaping Africa’s AI Future
What Will It Take to Make AI Work for African Languages?
As artificial intelligence continues to develop, an important question remains: are African languages receiving the attention they deserve?
The African Languages AI Safety Readiness research highlights the need to make language inclusion a priority in AI development.
For researchers, one important responsibility is to evaluate how well language models perform across different languages and cultures. Quality evaluation helps identify weaknesses, biases and inaccuracies that might otherwise go unnoticed.
Governments also have a role to play. National AI policies should prioritise language inclusion, provide dedicated funding and create opportunities for local researchers and developers working on language technologies.
Making African languages part of AI development will require contributions from governments, local communities and research institutions.
As AI becomes more widely used, ensuring that African languages are represented is not something that can be left until later.
Read the research: African Languages AI Safety Readiness Report
Can Nigeria’s Public Institutions Keep Up with AI Adoption?
Artificial intelligence is becoming increasingly relevant to public services. From healthcare and education to security and revenue generation, its potential applications are growing.
However, adopting AI within government institutions comes with challenges that extend beyond the technology itself.
In an August 2026 study published in the International Journal of Public Administration, researcher Ubokuloh Okon examined the opportunities and challenges of AI adoption in Nigeria’s public sector.
The study highlights several areas where AI could support public institutions, including healthcare delivery, education, security, revenue generation and evidence-based policymaking.
However, it also identifies major barriers:
- Weak regulatory frameworks.
- Inadequate digital infrastructure.
- Shortage of skilled professionals.
- Cybersecurity threats.
- Data privacy concerns.
The researcher argues that Nigeria needs stronger institutional and legal frameworks, investment in digital infrastructure and human capital, and collaboration between government, academia and the private sector.
The findings highlight an important consideration for Africa’s AI ecosystem. Developing AI capabilities is only part of the challenge. The institutions, infrastructure and policies needed to support their responsible use matter just as much.
Read the research: Artificial Intelligence, Governance, and Public Policy in Nigeria
Nigeria Unveils National Digital Cloud Policy
Nigeria is taking another step towards strengthening its digital infrastructure with the unveiling of the National Digital Cloud Policy.
Introduced by the Federal Government through the Federal Ministry of Communications, Innovation and Digital Economy, the policy sets out plans to attract investment in cloud computing and data centres, strengthen local infrastructure and improve government digital services.
The policy focuses on four main areas:
- Attracting investment in cloud infrastructure.
- Positioning Nigeria as a regional hub for digital services.
- Expanding the use of cloud services across government institutions.
- Strengthening the protection of sensitive government and regulated data.
It also introduces measures to encourage local and international companies to invest in Nigeria’s cloud ecosystem. These include investment incentives, coordinated government procurement and support for developing local technical skills.
According to the Minister of Communications, Innovation and Digital Economy, Dr. Bosun Tijani, the broader goal is to move Nigeria beyond being mainly a consumer of global cloud services.
The ambition is to create more opportunities for infrastructure development, investment, technical skills and digital services within the country.
For Nigeria’s growing technology ecosystem, the policy represents an important conversation about digital sovereignty and the infrastructure needed to support future innovation.
Read the announcement: National Digital Cloud Policy
03. INNOVATION: Emerging AI Technologies and Solutions Across Africa
Across the continent, researchers, technology companies and entrepreneurs are developing AI solutions for language processing, digital communication and agriculture.
Some are building foundation models specifically for African languages. Others are exploring offline translation, speech recognition and digital platforms that support agricultural operations.
Here are five innovations demonstrating different approaches to applying artificial intelligence to African challenges.
MORENA: Building an AI Model for African Languages
Developed by: Vambo AI, South Africa
MORENA is an African language model built from scratch to support 12 African languages, alongside English and French.
Unlike approaches that adapt existing models, MORENA was developed with its own tokenizer and training data mixture designed to improve how African languages are represented and processed.
Vambo AI reports strong results on African-language text modelling and translation benchmarks, including outperforming several larger models in its evaluations.
Its model weights are released under the Apache 2.0 licence, allowing developers and researchers to use and adapt the technology.
Explore the model: MORENA on Hugging Face
CommonLingua: Recognising 334 Languages
Developed by: Pleias with the GSMA
Before an AI system can process text in a language, it must first identify that language correctly.
CommonLingua addresses this challenge through an open-source model that recognises 334 languages, including 61 African languages.
With approximately two million parameters, it is considerably smaller than many existing systems.
Rather than relying on a language-specific tokenizer, CommonLingua works directly on text bytes. This allows it to handle different writing systems and support the organisation of multilingual datasets at scale.
Explore the model: CommonLingua on Hugging Face
TranslatePsy-AfriSLM: Translation Without an Internet Connection
Developed by: Tether AI Research
Internet access is not always available, particularly in areas where reliable digital infrastructure remains a challenge.
TranslatePsy-AfriSLM is an open-source machine translation model designed to work directly on smartphones and computers without an internet connection.
It supports several languages, including Hausa, Amharic, Yoruba, Igbo, Swahili, Zulu and Somali.
Because translation happens locally on the device, users do not need an active internet connection to access its functionality, and their data stays on the device.
The developers report competitive results against substantially larger systems on selected benchmarks.
The technology also provides a foundation for developers interested in building translation applications for education, healthcare and other services.
Explore the technology: TranslatePsy-AfriSLM on Hugging Face
Longa: Bringing Farmers’ Voices into Agricultural Research
Developed by: IFPRI and Farm Radio International
Farmers regularly share their experiences, challenges and needs through radio programmes and voice messages.
However, analysing large volumes of recordings can be difficult, particularly when the conversations involve languages with limited digital resources.
Longa is designed to address this challenge by transcribing, translating and analysing farmers’ messages.
The project currently starts with Luganda, mainly spoken in Uganda, and Bambara, mainly spoken in Mali.
By making these conversations easier to analyse, the technology helps agricultural organisations bring farmers’ feedback into research and advisory services.
Read more: IFPRI and Farm Radio International’s Longa Project
AquaTrack: Bringing Digital Tools to Fish Farming
Developed as a Nigerian aquaculture platform
Fish farming involves keeping track of several activities, including feeding schedules, water quality, fish growth, inventory and expenses.
AquaTrack brings these activities together in a digital farm management system.
The platform provides tools for recording farm activities, monitoring performance and managing finances. It also offers AI-powered recommendations on feed quantities, water management and stocking densities.
While the platform presents an interesting application of technology in agriculture, independent evidence of its impact on fish yields and farm profitability remains limited.
Explore the platform: AquaTrack Management System
04. OPPORTUNITIES: Events, Jobs and Learning
For researchers, developers and technology enthusiasts looking to take the next step, here are some opportunities worth exploring.
International Conference on Next-Generation AI Technologies (ICNGAIT)
A two-day conference bringing together academic and industry perspectives on emerging AI technologies, research and applications.
Date: 10–11 November 2026 Location: Nairobi, Kenya
Senior AI Engineer: Agentic Systems and Automation
An opportunity for experienced AI engineers to design and deploy agentic AI systems, intelligent document pipelines and internal AI assistants.
Location: Fully remote from South Africa Application deadline: 7 November 2026
Tech Lead: Energy and Climate Intelligence, Wattzo
Wattzo is looking for a technology lead to build AI-powered energy and climate intelligence systems.
The work focuses on developing solutions that help utilities, cities and communities make smarter decisions.
Location: Remote-first with flexibility Application deadline: Not specified
AI Application Engineer: Co-operative Bank of Kenya
This role focuses on designing, developing and maintaining AI-powered features.
Responsibilities include working with retrieval-augmented generation (RAG), conversational agents, tool-calling workflows and intelligent automation.
Location: Nairobi, Kenya Application deadline: 9 November 2026
05. AI TRENDS AND NEWS: Developments Worth Following
Are AI Models More Understandable Than We Thought?
Understanding how AI models work and how their internal processes can be studied remains an important area of AI research.
This issue highlights a growing conversation around the interpretability of AI models and what researchers are learning about their behaviour.
Anthropic Warns of Potential AI-Powered Self-Replicating Malware
As AI capabilities advance, cybersecurity researchers are examining how these technologies could be used to develop more sophisticated forms of malicious software.
Anthropic researchers are investigating how AI could enable malicious software to spread across connected devices.
Their findings highlight a potential cybersecurity risk as AI systems become more capable.
However, a fully autonomous AI worm spreading across smart home devices has not yet been demonstrated.
The distinction matters. Potential threats deserve attention, but they should not be confused with attacks that have already been demonstrated.
Read the research: Anthropic’s Research on Self-Replicating AI
Denmark Moves to Protect Faces and Voices From AI Deepfakes
AI-generated images, voices and videos are becoming increasingly realistic, raising concerns about consent, identity and digital rights.
Denmark has introduced a bill that would make it illegal to share realistic AI-generated replicas of people’s appearance, voice or movements without their consent.
The proposal aims to give individuals greater control over their digital identities and provide stronger protections against harmful deepfakes.
As generative AI becomes more accessible, developments like this highlight the growing need for legal frameworks that address how people’s identities can be used and reproduced online.
Read more: Denmark Introduces a Bill Protecting People’s Faces and Voices From Deepfakes
06. THE ENCODER RIDDLE: Did You Crack the Code?
Before we wrap up, here is the riddle from this edition.
The answer is: DATA.
Data is what an AI model learns from. The quality, accuracy and availability of that data influence what a model can learn and how well it performs.
This also brings us back to one of the important conversations in this edition: does a bigger AI model always mean a better one, or does the quality of the data matter just as much?
For African languages and other underrepresented areas of AI research, this question is particularly important.
Building better AI systems is not simply about making models bigger. It also means paying attention to the data used to train them, whose experiences that data represents and what might be missing.
Until Next Time
The work of building useful AI systems continues across Africa. From research and education to language technologies, agriculture and digital policy, progress depends on people willing to explore ideas, test solutions and improve what already exists.
Keep learning. Keep researching. Keep building.
