Why an Imperial PhD candidate trusts herds of cattle more than satellites

In 2021, Kayode Adeniyi built a data collection app for Nigeria’s Ministry of Education to gather information on libraries across several states, from their facilities to the resources they held. His team tested the app, deployed it to the field, and waited for the data to come in.

Then the complaints started.

Data collectors in northern Nigeria were filling out the forms, but poor internet connectivity meant much of the information never made it to the database. Work that had been completed in the field was disappearing before the team could use it.

“We were building as if all the data collectors were in Garki or Wuse, where we had a good internet connection,” Adeniyi says. “If we actually scoped where we were designing this app for initially and we knew that there’s no internet there, we would have done it differently.”

The team eventually integrated SQLite, a serverless SQL database engine that stores an entire database in a single standard file. This allowed the app to store data locally and upload it when the collectors returned to an area with a stable connection.

But by then, they had lost weeks of work and had to return to the field to collect the data again.

For Adeniyi, the episode captured a problem he has encountered repeatedly in technology: a system can work exactly as it was designed and still fail the people who are meant to use it.

“These things are different on the ground,” he says.

Collecting farm-level data in Saminaka, Lere Local Government Area, Kaduna State, one household at a time. Image Source: Kayode Adeniyi

From Architecture to Geography

As a child growing up in Ado Ekiti, the capital of Ekiti State in southwestern Nigeria, Adeniyi wanted to become an architect. New buildings in the city made him curious about how structures were designed and how cities were put together.

After secondary school, he planned to study architecture at Ladoke Akintola University of Technology. Instead, in 2015, he chose geography at the University of Ilorin.

His interest in technology had started much earlier.

At Christ’s School in Ado Ekiti, where he attended secondary school, the government brought laptops for students. A teacher trained them to use the computers and introduced them to programming languages including Fortran, BASIC and COBOL.

Adeniyi became so interested in computers that he and a few friends would stay in the laboratory during siesta.

“Instead of observing the siesta period, we would be in the lab,” he says. “Practising and learning by building some games and apps.”

Some of the laptops eventually stopped working. At the University of Ilorin, however, geography gave Adeniyi another route into technology.

After his first year, a geography course introduced him to Geographic Information Systems (GIS). He learnt that mapping technology could be used to create maps instead of drawing them by hand.

Intrigued, Adeniyi stayed behind in Ilorin during a holiday, paid for extra classes and learnt to use Python and other software to make maps.

He soon began helping master’s students create maps for their projects and dissertations. They paid him for the work. Friends studying computer science and political science also encouraged him to learn Python properly. One of them downloaded Learn Python the Hard Way, a Python textbook by Zed Shaw, onto his phone.

“That was what got me into it,” he says.

From maps to systems

Adeniyi soon realised that coding could do more than help him draw maps faster. He began using Python and JavaScript on projects that showed him how geography could be applied to problems beyond the classroom.

For a professor at the University of Ilorin Teaching Hospital, Adeniyi mapped roads, population distribution and traffic patterns in Ilorin to determine where ambulances could be positioned to reach people within a five-kilometre radius.

He also worked on telecommunications distribution networks, studying how mobile phone masts could be positioned to serve surrounding communities. For his final-year project, he used satellite imagery to study the environmental impact of an underpass near the airport.

The projects gave him a clearer idea of where his skills could take him.

“Instead of me finishing from school and I cannot pinpoint where I want to work, but they are telling me that you can work anywhere; I already knew places I can work,” he says.

After graduating in 2019, Adeniyi joined Sambus Geospatial Nigeria Limited in Abuja for his mandatory one-year National Youth Service Corps (NYSC), where he worked as a GIS analyst.

After completing his service, he left Sambus and joined Lorex as a geospatial software engineer.

At Lorex, he worked mainly on prototypes for government-facing institutions, helping to build mapping technologies and public-sector data systems. Some projects helped states identify mineral resources and estimate how much remained, while others involved building data infrastructure for services such as routing.

The work also showed him that writing code was only one part of building a useful system.

“You spend, let’s say, 20% of your time writing code, building apps, integrating things, but you spend the bulk of the time cleaning the data,” he says.

**When tech is not enough **

In 2021, Adeniyi joined a third-party consultancy working on the Central Bank of Nigeria’s Anchor Borrowers’ Programme. The team registered farmers and helped ensure they received seedlings, fertilisers and payments.

The work brought together his geospatial experience, but also exposed him to a problem that technology alone could not solve.

“I saw how many people were still unbanked despite the flood of fintechs in the country,” he says.

The team was using satellite information and GeoAI to work with farmers, but Adeniyi saw that technology could achieve little if farmers could not receive the money they were entitled to.

“We were doing serious work with satellite information and GeoAI for agriculture, and none of it means anything if the farmers themselves are not getting paid,” he says.

In 2022, he joined Flutterwave as a full-stack engineer, working with the business development team to build custom enterprise payment solutions for small and medium-sized businesses.

The role exposed him to the complexity behind transactions that users often experience as simple debits or transfers. He worked with clients to design solutions across businesses and financial institutions.

Accra, Ghana for the company-wide Flutterwave retreat, out at Independence Square with the team. Image Source: Kayode Adeniyi

“I got to know how complicated payment is too,” he says. “You cannot see me complaining about my money getting debited and all that. Because I feel the pain of the guys on the backend that are putting things together.”

Working closely with the business development team also showed Adeniyi that engineering decisions were rarely made in isolation. Businesses had requirements, regulators had rules, and policy shaped what could ultimately be built.

“The biggest thing I learned there is that what gets built is shaped by policy,” he says.

From building to questioning

His growing interest in the forces shaping technology eventually took him to the London School of Economics and Political Science (LSE).

Adeniyi had initially planned to pursue an MBA in the United States to better understand the business side of technology. Personal circumstances changed those plans, however, and he chose LSE, where he studied Digital Innovation.

Kayode at the LSE LIFE section inside the LSE Library. Image Source: Kayode Adeniyi

The programme pushed him towards policy thinking and made him more interested in why digital products are designed the way they are, and what shapes those decisions.

“It’s a policy school. So you do more policy thinking, because the entire motto of the school is understanding the causes of things,” he says.

He began writing about technology and the systems around it, but he did not want to stop at analysis.

“There’s something they call a think tank,” he says. “So I prefer to be a do tank. As you are thinking, you can also do.”

His interest in artificial intelligence was also growing.

Before generative AI became mainstream, Adeniyi had already been using computer vision and machine learning to work with satellite imagery. He also volunteers with the GeoHazards Risk Mapping Initiative, where he works on modelling flood risks across Nigeria and Ghana.

One problem that interested him was cloud removal. Satellite images of much of Africa can be obscured by cloud cover, so Adeniyi worked on models designed to remove the clouds.

But removing too much could also erase the features researchers were trying to study.

“It is like trying to clean a photograph,” he says. “If you clean it too much, you remove the picture itself.”

His confidence in what he could do with AI grew through competitions and projects at LSE. He won the Phelan US Centre’s 2025 master’s students essay competition on AI, giving him the opportunity to present his essay to members of the UK Parliament.

Presenting the winning LSE essay on AI to MPs and peers of the British American Parliamentary Group. Image Source: Kayode Adeniyi

“When you go to hackathons, or you enter competitions, competing with maybe 200 other people and you win, it also sparks your confidence,” he says. “It makes you believe it’s possible.”

When the machines needed help

Adeniyi’s work with computer vision and satellite imagery also made him think more closely about the role of local knowledge in technology.

In Lere Local Government Area of Kaduna State in northwestern Nigeria, farmers were losing their crops to floods almost every year. While studying at LSE, Adeniyi began asking not only what a satellite or machine-learning model could predict, but what the people living with the floods already knew.

Farmers had their own ways of recognising when flooding was coming. One sign, Adeniyi learnt, was the behaviour of their cattle. When the animals began avoiding particular areas, farmers knew something was changing.

“The satellite will revisit like six to 12 days, but the herds, they’re always there 24 seven,” Adeniyi says. “So they are like better sensors, quote unquote, for our models.”

The observation changed the way he approached flood mapping.

Rather than treating local knowledge as separate from technology, he began using it to improve the models. People on the ground could provide landmarks and identify areas that had flooded before, while satellite imagery and machine learning could map those patterns at a larger scale.

“Technology itself can hinder development, but when it is now merged with local knowledge and crafted with consciousness, you can get better results,” he says.

At LSE, Adeniyi also began exploring another application of AI: using machine learning to analyse qualitative social science research at scale. The project examined how well large language models could handle information that researchers usually have to interpret themselves.

The more he worked with these models, the more interested he became in how they handled cultural meaning.

“Because fundamentally, words are very important, even cultural context,” he says.

That interest led him to another problem: how to make disaster warnings understandable to the people receiving them.

Adeniyi designed a system that uses natural language processing to convert disaster information into local languages such as Hausa, Yoruba, Efik and Ijebu. But language was only part of the problem. The system also needed to communicate location in ways people could understand.

“Then let’s say you now embed that into the 3D vision on Google Maps, you can begin to describe places to them, like beside the mosque, beside the church, on the bridge, towards the farm,” he says. “So the description will not be too abstract to them.”

Adeniyi says the system is now being used as a case study in Kacha Local Government Area of Niger State, where he and his collaborators are working with farmers growing flood-resistant rice and providing information in the local language they understand.

That work connects to a bigger question that now interests him: how much confidence should people place in what a machine produces?

Adeniyi calls it uncertainty quantification.

He explains the idea through mapping. Canada sits farther north, where the Earth’s surface is represented differently by map projections than in Nigeria, which is closer to the equator. As a result, five feet measured using one projection may not represent exactly the same amount of ground as five feet measured using another.

A system can perform perfectly correct arithmetic on those numbers and still produce the wrong answer.

“The maths will be perfectly correct, the answer will still be wrong,” he says. “And the most dangerous part is that the error is completely invisible.”

For Adeniyi, uncertainty quantification is about making that invisible error visible and accounting for what a system cannot reliably do.

The next frontier

That question is also behind his choice of research for the next stage of his career.

In September, Adeniyi will begin a fully funded PhD in Engineering at Imperial College London, where he will research machine-learning methods for designing hardware sensors for global water risk.

The problem, he says, is that the world is running out of fresh water, yet researchers do not have a clear picture of how much remains or where it is. Better sensors, combined with machine learning and satellite information, could help measure those resources and quantify the uncertainty around them.

“If you unlock this first layer, then we can do other things,” he says. “Get the uncertainty quantification on that area, then we can begin to do other things with water.”

Alongside the PhD, Adeniyi plans to continue working on AI evaluation and threats, particularly how organisations can assess the risks of increasingly capable AI systems.

“How do we evaluate the threats? How do we evaluate the risk of all those things in organisational usage?” he says.

For Adeniyi, the lesson from years of building technology is straightforward: the people who will use a system have to matter before the system itself.

“Before you sit down to design and to build, talk to people,” he says. “What you think is important might not be the important thing at all to them. They might not even want it. And being open-minded comes with that too.”

True scale demands moving beyond surface-level integrations to robust execution. We’ve filtered the noise out of Moonshot 2026, optimising the conference strictly for high-calibre connections between startup founders, global financial operators, enterprise leaders and individuals rewiring Africa’s technical frameworks. Get 20% off Early Bird tickets for a limited time.