On the southern fringes of Kenya, in the shadow of Mount Kilimanjaro, a boy once dreamed of becoming a footballer. But a diagnosis of myopia quietly ended that pursuit.
Medicine, the path his father favoured, was never an option, he couldn’t stomach the sight of blood. By the time Stephen Kolesh walked into the dimly lit computer lab at Oloolaiser High School, he was a teenager with no clear ambition, until he wrote his first line of HTML code.
Kolesh, now one of Africa’s most accomplished machine learning competitors, was born in Loitokitok, a small town in Kenya’s Kajiado County. His childhood was relatively stable, but it is precisely this ordinariness that makes his rise so remarkable.
“I honestly never knew what I wanted to be,” he recalls. “My parents wanted me to pursue medicine since my dad is a doctor, but I had a phobia of blood, so that path never appealed to me.”
A Spark in the Computer Lab
Kolesh’s journey into technology began in the computer lab at Oloolaiser High School. “During club days, I would spend my time in the computer lab teaching myself web development using HTML, and I found it incredibly fun. From that point on, I knew I wanted to do something related to computers,” he says. The spark ignited there would soon become a blaze.
After high school, a long break before university led Kolesh to binge-watch the TV series “Mr. Robot,” a show about a vigilante hacker. The protagonist, Elliot Alderson, inspired him to pursue a Bachelor’s degree in Computer Science. “I was so inspired by the character that I decided to pursue a Bachelor’s degree in Computer Science,” Kolesh says.
At Multimedia University of Kenya, Kolesh’s path took another turn. During orientation, a lecturer declared that data science would be “one of the sexiest jobs in the next 10 years.” Kolesh admits, “I had never even heard of data science before. At the time, all I wanted to be was an ethical hacker inspired by Mr. Robot. That one statement sparked my curiosity, and I started researching what data science was all about.”
He found his tribe in the university’s Computer and Information Technology (CIT) Club, which hosted daily tech sessions. “If I remember correctly, the data science sessions were held on Tuesdays, and I made it a point to attend them regularly,” he says. There, he met Lawrence Morueye, a student who had won the UmojaHack hackathon—one of Africa’s largest university competitions in data science and machine learning. “I thought that was incredibly cool, and it only fueled my curiosity and passion even more.”
Enter Zindi: A Continent-Wide Community
Kolesh’s introduction to Zindi, Africa’s leading data science competition platform, came through the CIT Club. “The fact that someone from our university—someone I had actually met—had already won multiple competitions on the platform really inspired me. That’s what motivated me to participate in my first challenge; I wanted to be like him. Plus, the idea of winning a competition across all of Africa sounded so cool. I wanted that for myself too.”
From his very first competition, Kolesh was hooked. “The competitions Zindi hosted were based on real-world problems affecting the African community, and the solutions developed by participants were actually being implemented to address those issues. That’s when it hit me—these weren’t just toy problems or practice exercises. They had a real impact. It was at that moment I realized I should take Zindi seriously.”
Kolesh’s technical growth on Zindi has been nothing short of exponential. “All competitions are technically challenging in their own way, but the most difficult ones for me have always been those in fields I hadn’t explored before,” he explains. He recalls the Makerere Passion Fruit Detection challenge as a formative experience.
“At the time, I had never worked on a computer vision project—my experience was limited to tabular data competitions. I was still in my first year of university, and we hadn’t even covered advanced topics like neural networks yet. To make things even harder, there was no ChatGPT back then to guide me. So, I had to rely on reverse-engineering existing solutions I found on GitHub. It was a tough learning curve, but that experience taught me a lot.”
Over five years on Zindi, Kolesh has amassed an extraordinary record: 38 gold medals, 17 silver, and 14 bronze. “In general, competitions that introduce me to completely new fields tend to be the most technically challenging, but also the most rewarding,” he says.
Building for Impact: AI in Disaster Response
Of the many competitions Kolesh has entered, none has resonated more deeply than the Kuyesera AI Disaster Damage and Displacement Challenge. The goal: develop a machine learning model to identify the locations of houses and determine whether they had been damaged by Cyclone Freddy.
“What made this competition especially meaningful was that it tackled a real humanitarian issue—providing rapid, data-driven insights in the aftermath of a natural disaster. The potential impact of such a solution in helping coordinate relief efforts and assess damage on the ground was massive,” Kolesh says.
The technical demands were steep. “I explored advanced architectures like MambaBDA, trying to push the performance of the model as far as I could. Even though that particular approach didn’t yield the best results, I learned an incredible amount through the process. In the end, we settled on a combination of YOLO for object detection and a separate image classifier to improve accuracy. It was a tough challenge, but one that left me with both practical skills and a deeper sense of purpose in using AI for social good.”
Zindi has been more than a proving ground for Kolesh’s technical skills—it has been a launchpad for professional growth and community building.
Stephen says that technically, Zindi exposes him to real-world problems across various domains such as computer vision, natural language processing, and geospatial analysis—experiences that help him develop a broad skillset beyond what he learns in the classroom.
He explains that he teaches himself new tools and frameworks like YOLO, XGBoost, and segmentation models by applying them to practical challenges. Over time, he grows more confident in reading research papers, experimenting with different architectures, and building end-to-end machine learning pipelines.
Professionally, he notes that the platform gives him visibility and valuable connections. He meets collaborators and mentors, and learns how to properly document his work—skills that prove useful in securing internships, contributing to open-source projects, and performing well in interviews. He adds that working on impactful challenges and competing at a continental level boosts his confidence and reinforces what he is capable of achieving.
Recognition on the Global Stage
Kolesh’s achievements have not gone unnoticed. In 2023, he won the Carbon Dioxide Prediction Challenge at UmojaHack Africa, the continent’s largest machine learning hackathon, beating more than 1,000 students from 345 universities in 36 countries.
“Ranking highly and being acknowledged on global platforms validated my efforts and made me realize that I could compete and contribute at a global level,” he says. “These milestones shifted my mindset from just wanting to be good at machine learning to actually wanting to make a real impact through AI, especially in solving African problems.”
Kolesh’s story is emblematic of a broader movement. Zindi, which now connects nearly 350 higher education institutions and thousands of students across Africa, is nurturing the next generation of AI talent.
Kolesh is now a mentor and role model for aspiring data scientists across the continent. His advice is pragmatic and rooted in experience: “Start with what you have, and stay consistent. When I began, I didn’t have access to fancy equipment, expensive courses, or even advanced classes in school. But what I had was curiosity and the internet. Platforms like Zindi, Kaggle, and YouTube, along with free resources like Fast.ai, and Coursera deep learning specialization, can teach you so much if you commit time to them.”
He urges young people not to wait until they feel “ready” to start building and competing. “Jump in early. You learn the most by doing. Even if your first few attempts feel like failures, every challenge teaches you something new. That’s how I learned—by struggling through things I didn’t know and figuring them out bit by bit. You just need grit, curiosity, and the willingness to keep learning.”
Shaping the Future of Global AI
Looking ahead, Kolesh believes Africa’s data scientists are poised to lead, not just participate, in the global AI revolution. “Africa has unique challenges, diverse data, and untapped perspectives that the world often overlooks. Solving problems here requires creativity, resilience, and deep contextual understanding—qualities that make African data scientists uniquely positioned to build inclusive, scalable, and globally relevant AI solutions.”
He sees a future where African voices are central to the ethical and responsible development of AI worldwide. “As the world moves toward more ethical and responsible AI, voices from underrepresented regions like Africa are needed more than ever to ensure AI works for everyone—not just a few.”
Kolesh’s personal mission is clear: “I want to build AI systems that solve real-world problems across healthcare, agriculture, disaster response, and education—not just in Africa, but globally.”
For Kolesh, the lesson is simple but profound: “You just need grit, curiosity, and the willingness to keep learning.”
With every challenge he tackles on Zindi, Kolesh is proving that brilliance doesn’t have a zip code. It needs only opportunity, community, and the courage to try.
Stay ahead in the world of AI, business, and technology by visiting Impact AI News for the latest news and insights that drive global change.
