Building reliable AI agents: The Bukola Sobowale example

During the early commercial application of machine learning (ML) and artificial intelligence (AI), especially in the last ten years, most of the current users of these rapidly evolving technologies either condemned them or were sceptical about their applicability. On the other hand, there were a few who believed, and still believe, that these technologies are the lifeblood of any innovation anyone could ever build and that the life of modern man and society depends on them.

In the last decade, it is seemingly impossible to think of anything worthwhile that was produced without the creative use of AI and other related technologies. Yet, in AI applications, one aspect that is often overlooked is the human side of the tool — the people behind the advancements, the decision-makers, the visionaries who shape its future, and the narratives that offer a critical understanding of what it takes for it to succeed.

During the invention of some of the recent chatbots, virtual assistants, and predictive analytics AI tools in use today, one of Nigeria’s young software and technology experts, Bukola Sobowale, was still studying Microbiology at the Obafemi Awolowo University, Ile-Ife, Osun State, Nigeria.

Sobowale revealed that, while studying Microbiology, he was first introduced to computer programming through a year-long Java course and later, during his third year of study, his interest in data science, artificial intelligence, and machine learning, motivated him to pursue a second degree in Electrical and Electronics Engineering at the University of Ibadan to deepen his technical foundation in and software engineering and AI technologies.

In recent years, in the course of applying his software and AI expertise to understanding human and societal needs, Sobowale stated that he has ventured into some innovations that can help people, businesses, and organisations in many ways. He added that value creation, using technology and his understanding of societal digital needs, has been his clarion call since he knew the immense power and opportunities that machine learning and AI could offer at his beck and call.

In the past few years, he has worked on projects across edtech, blockchain, and environmental solutions. In edtech, he contributed to Cognota, where he was part of a global team that built a platform to help enterprise teams manage and streamline internal learning programmes. In the blockchain space, he has worked on Algodex, one of the first decentralised Automated Market Maker (AMM), which facilitates trading of Algorand assets and other tokens with a focus on speed, low fees, and security.

He has also worked with OffBlocks, a United Kingdom-based blockchain company that provides a decentralised payment network and a developer toolkit that enables web3 developers to issue compliant bank accounts and debit cards, which brings a real-world financial infrastructure into the blockchain ecosystem.

“On environmental solutions,” he added, “I was a lead and founding engineer at Pakam, a platform that connected informal waste collectors to recycling companies using real-time logistics and geofencing, promoting sustainability and accountability in waste management.”

Crime and insecurity across Nigeria have been recurring issues that have proven difficult to curb. The traditional ways of managing these challenges have failed, forcing security agencies, policymakers, businesses, governments, and even local communities to turn to AI for solutions.

Sobowale’s interest in public and national security led him to recently invent a smart weapon detection system using a Raspberry Pi connected to a CSI camera and a custom-trained YOLOv8 CNN model, which he trained using datasets from Roboflow and optimised on Google Colab using NVIDIA SMI GPU instances. According to him, the system could identify concealed and visible weapons from live camera feeds and immediately send SMS alerts to security personnel via the Twilio API. This, he added, is particularly useful for security in schools, churches, and event venues.

“The idea stemmed from the increasing insecurity in public places, where traditional surveillance falls short. I wanted to create a smart, cost-effective solution that could be deployed even in resource-limited environments,” he said.

Another of his technological innovations, he revealed, was a smart water irrigation system that uses a machine learning model to decide when and how much water crops need based on weather forecasts and other environmental sensors such as humidity, temperature, soil moisture, among others.

He added, “The model was trained on historical weather data and plant watering patterns to reduce water wastage in small-scale farms.”

Perhaps, Sobowale’s most interesting AI-powered innovation that would interest common people is the context-aware WhatsApp bot with AI capabilities. According to him, this memory-enabled WhatsApp bot uses whatsapp-web.js, enhanced with Openai’s GPT API and a lightweight memory layer to maintain context across conversations, which allows the bot to deliver personalised and consistent responses over time, recalling previous chats, user preferences, and intent.

He revealed that the bot serves as a virtual assistant, capable of answering frequently asked questions (FAQs), pulling customer-specific logistics data, and handling real-time queries autonomously.

“This bot played a crucial role in improving customer support operations for an agricultural company, handling over 70 per cent of customer interactions without human intervention and significantly reducing response time,” he said.

The essence and usage of all AI tools and software is to make life and work easier. Bukola Sobowale, in a recent press statement, revealed how AI has made his work and inventions easier. One such way, he stated, was the application of faster prototyping with pre-trained models. He stated that pre-trained models like YOLO and GPT have drastically cut down the time it takes him to build intelligent systems.

“For instance, my weapon detection project would have taken months if I had to build a convolutional neural network from scratch. With YOLOv8, I had a high-performing model up and running in hours, ready to be customised and deployed with minimal training effort,” he said.

He further said another way that AI is making his work easier is through smarter coding with AI code assistants. He noted that AI tools like GitHub Copilot and Tabnine have significantly boosted his coding speed and accuracy. He further noted that whether he is scaffolding boilerplate, generating unit tests, or writing API documentation, these assistants help to reduce context switching and eliminate repetitive work.

“They also serve as real-time pair programmers, offering relevant code suggestions that align with best practices, making it easier to focus on problem-solving and architecture,” he added.

The AI expert and technology entrepreneur also revealed that streamlined data labelling and training pipelines are other ways that AI is making his innovations easy. According to him, with tools like Roboflow and synthetic data generators, preparing high-quality datasets is now faster and more scalable. He said he used them extensively for the weapon detection system to generate image variants, augment the dataset, and label objects efficiently.

“This not only saved time but also improved model robustness by ensuring it could handle diverse scenarios in real-world environments,” he said.

AI is the hope and lifeblood of the future, and people like Bukola Sobowale know it. In light of this, when asked what he hoped to achieve soon with AI, he said his focus is on building AI systems that seamlessly integrate into traditional enterprise workflows, especially in critical sectors like healthcare, logistics, and fintech, where trust, accuracy, and transparency are non-negotiable. He emphasised that he is committed to developing reliable, real-time AI agents that do not just automate tasks, but also enhance human decision-making with explainable, responsible intelligence.

“In the short term, I am working on reducing hallucinations in retrieval-augmented generation (RAG) systems to make AI outputs more grounded and trustworthy. The ultimate goal is to democratise AI, embedding it into legacy systems and real-world infrastructure in a way that drives inclusive innovation and meaningful progress across the globe,” he said.

Sobowale stressed that with all the power, potential, and opportunities that AI and machine learning could offer the world and that he has personally enjoyed, he confidently could assure individuals, businesses, and governments that AI is not only here to stay for good, but is also a good force, and not a threat to people, jobs and industries.

The AI expert advised that instead of seeing AI as a competitor, it should rather be been seen as a collaborator, adding that the real advantage goes to those who learn how to work alongside AI, not against it.

“If AI can handle half of your tasks, then learning to leverage it can double your impact. When you understand both its strengths and limits, AI becomes a powerful ally, not a threat,” he advised.

As AI and related technologies continue to influence people and jobs, and to shape the future of industries, it is important to understand that their true potential lies not just in code and data, but also in the people who drive their development, application, and future.

Sobowale revealed that mentoring and coaching young software engineering students, technology enthusiasts, and AI innovators in their early careers is an endeavour that gives him joy. He stated that beyond mentoring on platforms like ADPList and JoinVeet, and engaging with communities on Slack and Discord, he also contributes as a panellist at technology events, where he speaks on building with emerging technologies and the future of AI.

He said that his mentorship style is hands-on and project-driven, encouraging mentees to bring real-world problems they care about and guiding them through the full AI development lifecycle — from data collection and model selection to deployment and post-launch evaluation.

“I regularly share practical resources and code snippets, and also help mentees validate their ideas with feedback grounded in industry-based cases. Whether on stage or one-on-one, my goal is to make cutting-edge technologies accessible and actionable for the next generation of builders,” he concluded.

ALSO READ TOP STORIES FROM NIGERIAN TRIBUNE