By Felix Akinlabi

A Software engineer explains the human decisions behind algorithms that shape millions of digital experiences daily

Every swipe through a personalized playlist, every “recommended for you” product suggestion, and every targeted social media feed represents hours of invisible engineering work happening behind the scenes. As artificial intelligence increasingly shapes how Nigerians discover content and make purchasing decisions, one Lagos-based developer is pulling back the curtain on what really makes these systems tick.

Oluwaseun E. Folorunsho, a software AI developer specializing in personalization systems, says the technology most people experience as seamless magic is actually the result of countless human decisions, complex engineering challenges, and careful ethical considerations.

“Personalization isn’t about reading minds, it’s about patterns,” explains Folorunsho, whose work involves building the infrastructure that powers recommendation engines for millions of users. “But before any recommendation is made, there are enormous amounts of invisible work happening in the background.”

The Hidden Infrastructure of Digital Recommendations

Behind every “You might also like” suggestion lies a complex web of data processing that most users never see. Folorunsho describes spending her days building pipelines that clean and transform millions of user interactions, clicks, purchases, searches, and viewing patterns into signals that machine learning models can interpret.

“We’re constantly making decisions about what behavioural patterns actually matter,” she said. “Does the time someone spends watching a video count more than whether they finish it? These aren’t just technical questions, they’re choices that shape what people see.”

The technical challenges are particularly acute in Nigeria, where infrastructure limitations add layers of complexity. “A system that works perfectly for 100,000 users can completely collapse when Burna Boy drops a new album and millions log in simultaneously,” Folorunsho noted. “We have to account for unreliable internet, expensive servers, and frequent power outages/realities that don’t exist in Silicon Valley.”

Human Choices Shape Algorithmic Outcomes

One of the biggest misconceptions about AI personalization, according to Folorunsho, is that it operates autonomously. In reality, every algorithm reflects deliberate human decisions about
what success means and what content gets promoted or filtered out.

“We decide whether the goal is to increase clicks, extend user engagement, or help people discover new content, ” she explained. “Each of those objectives produces dramatically different recommendations.”

These decisions carry significant responsibility. A music app might filter explicit content for younger users, while an e-commerce platform could avoid recommending out-of-stock items. But the implications go deeper than simple content filtering.

“Every time a personalization model goes live, it subtly reshapes the choices users make,” Folorunsho said. “We can introduce someone to a life-changing book or a new artist, but we can also reinforce echo chambers and limit exploration.”

Fighting Bias in the Nigerian Context

Addressing algorithmic bias presents unique challenges in Nigeria’s diverse market. Training data often skews toward urban users in Lagos and Abuja, potentially excluding rural consumers from receiving relevant recommendations.

“If our models are trained primarily on data from city users, they might completely fail to serve rural customers,” Folorunsho explained. The solution requires building flexible systems and constantly auditing for bias across different demographic groups and geographic regions. Her team collaborates with domain experts, ethicists, and social scientists to understand potential downstream effects.

“A model can be technically ‘accurate’ but still be unfair,” she noted. “Sometimes we deliberately choose a less accurate model if it proves more equitable in practice.”

Team Effort Behind Individual Experiences

Contrary to popular images of solitary programmers, AI development especially in personalization is intensely collaborative. Folorunsho works alongside data scientists who analyse user behaviour, product managers who define business objectives, frontend teams who build user interfaces, and domain experts who understand cultural context.

“I’ve collaborated with doctors on healthcare projects, economists on financial modelling, and linguists on natural language processing,” she said. “The best solutions come from combining technical capabilities with domain expertise.”

In Nigeria’s fintech sector, this collaboration extends to privacy experts and behavioral psychologists, ensuring that recommendation systems respect both user preferences and financial trust.

Looking Forward: Ethical AI Development

As personalization technology advances across Nigeria, Folorunsho sees both opportunity and responsibility. Improved mobile access, affordable data, and scalable cloud services are making AI systems more adaptive and locally relevant. “We’re not just optimizing for clicks or conversions,” she emphasized. “We’re designing algorithms that must be equitable, transparent, and worthy of users’ trust.”

For aspiring AI developers, Folorunsho recommends building strong foundations in programming and statistics while also studying economics, psychology, and ethics effective AI developers understand human behavior, business dynamics, and societal implications,” she said.

“The future of AI isn’t about faster algorithms or bigger datasets, ” Folorunsho concluded. “It’s
about building systems that are genuinely beneficial, ethical, and trustworthy. For those of us
working behind the scenes, we’re not just building systems that predict what people want, we’re building systems that influence how people discover, connect, and grow.”

As Nigeria’s digital economy continues expanding, the work of developers like Folorunsho
increasingly shapes how millions of people experience the internet. Their decisions about data,
algorithms, and ethics don’t just power recommendation engines, they help determine what kind
of the digital future the country will inhabit.