Welcome to Issue #51 of TAIS, where every Friday we spotlight visionary changemakers reshaping Africa’s tech, data, and AI landscape, one breakthrough at a time.
In today’s issue, we spotlight Maximillian Ntabo, a Kenyan AI/ML Solutions Architect and technical leader whose work sits at the intersection of engineering rigour and African context, where AI systems are either built to function in the realities that actually exist on the ground, or designed for environments that don’t.
Her world spans multiple fronts: architecting and deploying end-to-end machine learning systems for African businesses, leading data and AI strategy at Luanno Analytics, building MIP Intelligence, a platform designed to give African brands the kind of ROI visibility and marketing intelligence that has long been available to their global counterparts; and helping shape one of Nairobi’s most grounded builder communities through AI Tinkerers. Running through all of it is the single defining conviction that meaningful AI is not measured by the sophistication of the model but by whether the system actually works for the people it was built for, in the conditions they actually face.
Maximillian Ntabo | Building @MIP Intelligence | Data & AI|ML Engineer | Consultant
The conversation covers technical ground(architecture, deployment, low-resource systems, ethical practice) but it doesn't stay there long. It keeps returning to the same underlying question: what is African AI actually for, and who is it actually serving?
This is a story about a practitioner whose most important engineering decisions happen before any code is written; in the willingness to sit with the problem long enough to understand it, to design for the constraints that exist rather than the ones that would be more convenient, and to measure success not by what the system can do but by what it delivers to the people who depend on it.
Origin & Conviction
Q: You work at the intersection of AI engineering, data science, and technical leadership. What early influences shaped your journey into building AI systems for real-world impact?
A: My journey into AI and data science started with a simple desire: I wanted more for Africa. I believed that if we could build better systems that truly understand our context, we could help businesses grow and create meaningful impact. I was born and educated in Kenya under the 8-4-4 system, and like many people, I didn’t begin my academic journey pursuing what I was deeply passionate about. Most parents simply want their children to be employable, especially in a continent facing high youth unemployment, and I don’t blame them for that. Still, I knew early on that I wanted to follow my convictions. That decision wasn’t easy. At the time, Gen AI and advanced deep learning frameworks were barely adopted in Kenya, and I faced discouragement even from people already working in the field who told me there was no point in going further. But I chose to trust my instincts and commit to learning anyway. I dedicated myself to mastering data science, machine learning, and AI through self-directed learning, supported by formal certifications. This journey would not have been possible without the unwavering support of my family, and above all, my faith in God, which kept me grounded and resilient during moments of doubt. Being largely self-taught reshaped how I view education. I learned by building creating projects, making mistakes, and solving real problems for real businesses across different industries. Looking back, I’m grateful I followed my convictions. That choice-built resilience, discipline, and clarity of purpose. Today, it guides how I build AI systems: practical, responsible, and designed for real-world impact especially within the African context.
Editorial commentary: “I wanted more for Africa” is a deceptively simple sentence. Most readers will pass over it as a statement of motivation, but it quietly reveals how her journey into AI began. Technology does not appear first in her story. She does not describe discovering AI and then looking for problems to solve. Instead, she begins with a continental aspiration and only later arrives at AI as the instrument. That order is unusual. It suggests an entry point into the field that begins not with fascination for the technology itself, but with a question about Africa’s future with the technical pathway emerging afterwards.
There is another subtle signal in how she establishes credibility. She does not begin with institutions, titles, or employers. Her story starts with conviction, family support, and faith and only then moves to technical mastery. In many mainstream technology narratives, authority flows outward from credentials. Here it moves in the opposite direction: personal conviction first, expertise later. In emerging ecosystems, legitimacy often forms this way, not through institutional validation, but through persistence long enough for the work to speak.
It is a pattern that appears frequently in TAIS conversations. For many of them, the motivation arrives before the infrastructure. The conviction precedes the field itself.
Technical Architecture & Engineering Practice
Q: As an AI/ML Solutions Architect, how do you approach breaking down complex problems into clear, actionable engineering solutions, especially in African contexts where constraints are very real?
A: When I started working in Data & AI, my focus was simple: I wanted to build systems that actually make a difference in Africa. Early on, I realized that the key wasn’t just the technology it was understanding the problem first. What are the real challenges businesses face? How do constraints like infrastructure, language, and budgets shape what’s possible? That’s why I spent time researching and validating solutions before building anything. For example, while working with AWAN Afrika, which represents 44 countries across the continent, I dove into the regional differences, spoke with stakeholders, and tested assumptions to make sure the solutions we were designing actually made sense in each context. That effort led to functional AI prototypes like multilingual RAG systems for education and agriculture that weren’t just concepts, but real tools people could use. I approach system design the same way architects approach buildings: I start with a clear diagram. It brings order, helps me break down complexity, choose the right tech stack, and plan trade-offs all before writing a line of code. I applied this approach in startups like RMGDrive, building an AI-powered KYC verification system, customer retention engine, and Agentic AI RAG platform projects I handled end to end, from design to production. Today, I carry the same principles into MIP Intelligence and previously at Luanno Analytics though located in the UK, building AI products that serve African businesses, solving real problems, and delivering measurable impact. For me, the story of AI isn’t just about models or data it’s about creating solutions that work in the real world, in the places where they matter most.
This problem-led, context-aware approach inspired MIP Intelligence a platform being designed to address recurring gaps African brands face, from fragmented data to limited ROI visibility.
Hosted by Adamur, this end-of-year mixer highlighted the convergence of AI and Web3, emphasizing that meaningful impact and purposeful building in areas like agentic AI and blockchain infrastructure must always take precedence over hype.
Q: Walk us through your process for architecting and deploying an ML system end-to-end. What frameworks, tools, or principles guide your decisions from prototype to production?
A: My approach to architecting and deploying ML systems is problem-led and system-oriented, beginning with a clear definition of the business constraints such as latency, scale, cost, and data availability that the system must support. I transition from lean prototypes to production-grade systems by prioritizing modularity, reproducibility, and observability, using a toolkit tailored to the specific data and scale requirements. For instance, I use Pandas and NumPy for standard tabular EDA, PySpark for distributed big data processing, and OpenCV or Torchvision for image-based workloads.
A practical application of this was my work for Awan Afrika, where I developed a plant disease detection system using a ResNet-50 backbone to handle the high variance of real-world photos. Although the project served primarily as a technical validation and feasibility study due to the organization’s shifting capital priorities, I built the prototype with production-ready standards, focusing on handling “noisy” inputs and ensuring future scalability. When hardening such systems for deployment, I remove the “Python bottleneck” by exporting models to ONNX or TorchScript for C++-backed execution and utilize FastAPI for its high-performance, asynchronous capabilities.
To ensure efficiency in resource-constrained environments, I employ optimization techniques like transfer learning (like Awan Afrika), quantization, and pruning. Ultimately, my framework choices are driven by business outcomes rather than personal preference, ensuring that every system whether a full-scale deployment or a feasibility study is maintainable, auditable, and resilient to performance or data drift with monitoring in place to ensure reliability over time. Applying these same design principles, I’m now building MIP Intelligence, an AI platform that helps African brands measure ROI, optimize marketing spend, and gain actionable insights designed to scale even in low-resource environments.
Q: Many African founders face challenges around infrastructure, compute, and data quality. How do you design AI systems that remain robust in low-resource environments?
A: I’ve had to design for these constraints in practice. At RMGDIVE, the founder was very clear that resources were limited I remember being handed a Lenovo machine with a single GPU and being told, “this is what we have.” So, the system was designed around that reality. We used a hybrid approach: local compute where it made sense, cloud GPUs only when necessary, and CPU optimization for inference to keep costs down. Instead of training from scratch, I leveraged pretrained models and fine-tuned them for the specific use case automated KYC verification and customer support so we could ship faster and more efficiently. The models were treated as services, exposed via APIs, which made them easier to maintain, scale, and swap out as the startup grew. That combination hybrid infrastructure, model reuse, and service-based design allowed the system to remain robust, cost-efficient, and scalable even in a low-resource environment.
An Innovation manager meeting hosted by Qhala, this session explored why innovation often stalls at the boardroom level not because ideas lack value, but because they’re framed incorrectly.
Editorial commentary: Maximillian’s answers demonstrate a way of thinking about engineering that is rarely spoken aloud. For her, constraints define the shape of the system itself. Scarcity, uneven infrastructure, messy data, and shifting budgets guide every design decision from the very beginning. This approach builds foresight into the system. Prototypes are treated like full-scale systems, feasibility tests are designed to last, and hybrid architectures allow flexibility long before scale is guaranteed. The environment becomes a co-designer and technology a tool that responds to real-world conditions.
There is a subtle tension here that is easy to miss. Globally, AI often works the other way around; the model comes first, and the problem is fitted to it. For Maximillian, the sequence is inverted. The world comes first; the system follows. This reflects a distinctly African approach that is careful, adaptable, and rooted in reality, where credibility comes from getting systems to work reliably in the conditions that actually exist. The insight is clear, some of the continent’s most lasting technological solutions will emerge from engineers like Maximillian, people who can work within constraints, understand reality, and design systems that carry the logic of the environment itself.
Strategic Leadership & Industry Work
Q: You also serve as Director of AI & Technology at Luanno Analytics. What does this role involve, and how do you balance strategic leadership with hands-on engineering?
A: At Luanno Analytics, my role as a Data & AI Architect combined strategic advisory work with hands-on data engineering and analytics delivery.
A key client engagement involved improving how lead data was prepared, integrated, and used across CRM and marketing workflows to increase campaign efficiency and ROI. From a strategic standpoint, I worked with the team to understand the client’s commercial objectives, budget constraints, and outreach goals, ensuring that the data and analytics approach supported measurable business outcomes rather than just technical implementation.
On the execution side, I built and validated client datasets to ensure accurate CRM integration, reducing data errors and improving campaign reliability. I developed data segmentation strategies and lead scoring models to help prioritize high-value prospects and improve targeting effectiveness. I also created scripts and reusable frameworks for client tracking, segmentation, and long-term data hygiene, ensuring the solutions remained sustainable beyond initial delivery.
In addition, I supported the commercial side by advising on outreach strategies and budget allocation, helping clients fully realize their budgets through value-based pricing and optimized campaign design. I also contributed to writing proposals for prospective clients, clearly outlining how Luanno Analytics could deliver tailored AI and data solutions aligned to their specific business needs.
By operating across strategy, data architecture, and execution, I helped deliver clean, actionable insights that directly supported marketing performance and client decision-making.
Q: As someone who navigates engineering, community organising, and leadership, how do you ensure the AI systems you work on remain scalable, maintainable, and aligned with user needs?
A: I stay grounded in systems thinking and feedback loops. From an engineering perspective, I design AI as modular systems that can evolve without breaking, because scalability and maintainability are architectural decisions, not afterthoughts. From my community and leadership work, I stay close to users and builders, which keeps me honest about real needs versus assumed ones. Those inputs shape priorities, trade-offs, and when not to over-engineer. For example, with the MIP Intelligence platform, we’re currently planning scheduled market research across Africa. The goal is to validate assumptions, understand real user needs, and assess product-market fit before we build. This ensures that the architecture and ML solutions we develop are prioritized around actual impact, rather than hypothetical use cases. At the end of the day, alignment comes from continuously balancing technical rigor with human context, and building systems that can grow with their users, not ahead of them.
This approach guides the development of MIP Intelligence. Before scaling features, we conduct market research and gather feedback across East Africa to ensure the platform solves real problems for brands and agencies, not just hypothetical use cases
CTO End of Year Mixer Hosted by Qhala, focused on practical AI adoption in Kenya where real value is already being created through operational analytics and product innovation.
Editorial commentary: Reading between the lines of the work at Luanno Analytics, what becomes clear is that in applied AI, the smartest model or most elegant algorithm means very little if it doesn’t match the real-world needs of the organization using it. The work described isn’t flashy. It’s not the frontier model everyone reads about in tech news. But it’s the work that actually determines whether AI investments pay off, or end up as expensive tools nobody uses.
What stands out is the ability to operate across both strategy and execution. This is someone who can sit in a meeting about business goals and budgets, then go build the systems that make those goals possible. That combination is rare. Most organizations separate strategy from technical delivery. The gap between them is where most AI projects fail. A strategist who can’t build over-promises. An engineer who can’t translate business needs under-delivers. The people who can do both (translate business constraints into technical decisions and technical trade-offs into business implications) are the ones who make AI actually work.
Maximillian work shows more than skill. It reflects a philosophy about technical leadership in African AI contexts. Success depends on understanding both the technology and the realities of the organizations and markets it serves, and on building systems that work, not just models that impress.
Community Building & Ecosystem Development
Q: You’re deeply active in community-building, particularly through AI Tinkerers Nairobi. What gap did you see in the ecosystem, and how has a “code-first, no-pitch” culture changed the way AI builders collaborate?
A: As I became more active in Nairobi’s AI community, especially through AI Tinkerers Nairobi, I started noticing a gap in the ecosystem. There were many meetups and events, but most were heavily talk- or pitch-driven, which left little room for builders to actually experiment, learn, and collaborate on real code. What stood out to me about AI Tinkerers Nairobi was its “code-first, no-pitch” culture. By focusing on building rather than selling, it creates a space where people can show unfinished work, share experiments, and get practical feedback without pressure. That shift changes how AI builders collaborate people are more open, more curious, and more willing to help each other solve real technical problems. As someone who is actively attending these sessions, I’ve seen how this environment encourages deeper collaboration. Conversations move quickly from ideas to implementation, from theory to practice. Builders exchange approaches, challenge assumptions, and learn directly from what others are building. For me, being part of this community reinforces the importance of hands-on learning and shared problem-solving, especially in African contexts where practical, adaptable solutions matter more than polished pitches. It’s helping shape a more grounded, builder-led AI ecosystem.
Q: At AI Tinkerers, you recently highlighted projects ranging from career AI tools to IoT + Braille water-management systems. What do these projects reveal about the direction African AI innovators are taking?
A: At AI Tinkerers, we saw solutions ranging from career-focused AI tools to deeply contextual systems like IoT combined with Braille for water-management. That contrast is important it shows we’re not just experimenting with AI for novelty, but applying it across real socioeconomic layers: employment, accessibility, infrastructure, and sustainability. This is strong evidence that Africa doesn’t just have AI talent we have problem-driven innovators. Builders here are designing systems that are grounded in lived realities: unreliable infrastructure, water scarcity, inclusion for people with disabilities, and youth unemployment. These are not copied use cases from elsewhere; they are solutions emerging from African contexts. Beyond AI Tinkerers, I’ve also attended webinars focused on Africa’s AI readiness index, and what stood out was how prepared many ecosystems already are especially in terms of talent, community-driven learning, and early adoption. While we still face gaps in infrastructure and policy alignment, the innovation capacity is clearly present and accelerating. Overall, the direction is clear: African AI is moving from potential to practical impact. We’re building systems that solve African challenges, at African scale, with the capability to compete globally and that’s incredibly encouraging.
Sponsored by Enza Capital, the latest AI Tinkerers Nairobi highlighted a code-first, impact-driven culture where local builders prioritized practical applications and tangible execution over hype.
Editorial commentary: The distinction being drawn here between talk-driven and code-driven community culture sounds simple until you consider its downstream consequences for the shape of an innovation ecosystem. Across African tech hubs, the dominant format for community gathering has been the panel, the pitch, and the keynote, formats optimized for visibility, narrative, and investor-readiness rather than for the kind of technical depth and collaborative problem-solving that produces builders capable of going from prototype to production. The result, in many cities, is an ecosystem that is excellent at generating excitement about African AI and considerably less equipped to actually build it.
AI Tinkerers Nairobi represents, in this framing,more than just a different kind of meetup. It is a different theory of how technical capacity gets built at the community level. The “code-first, no-pitch” culture creates the conditions for something that polished-presentation culture actively suppresses: the willingness to show unfinished work. To share an experiment that didn’t work. To ask for help with a problem you haven’t solved yet. That willingness is the mechanism through which builders actually learn from each other, accelerate past individual limitations, and develop the kind of collaborative instinct that complex AI systems increasingly require.
The projects surfacing from this environment are also telling. They are not demonstrations of what African builders can do when given access to the same resources and frameworks as their counterparts elsewhere. They are evidence of what happens when builders who understand African problems deeply are given the space to define the solutions themselves. That is the difference between an ecosystem that produces talent and one that produces innovation. And it is a distinction that the continent’s AI investment community would do well to pay closer attention to.
Ethics, Responsibility & Impact
Q: When you evaluate the success of an AI project, whether in industry or community work, what metrics or indicators matter most to you?
A: For me, success is always about real-world impact. In projects like RMGDIVE’s automated KYC and AI-driven customer support, success meant reducing friction for users faster onboarding, smoother verification, and reliable support while also improving operational efficiency and scalability for the company. Similarly, with AWAN Afrika’s plant disease detection and farmer education platform, success was to be measured by tangible benefits to the end users: farmers being able to detect diseases early, get actionable recommendations, and ultimately increase crop yields. The system had to be accessible, easy to use, and cost-efficient so it could actually be adopted in the field. Across both cases, I look at user adoption, improved experience, and measurable impact whether that’s saving time, reducing errors, increasing revenue, or boosting yields. When AI delivers value to both users and the organization, that’s when a project has truly succeeded.
Q: Ethical and responsible AI are increasingly critical. In your work, how do you embed considerations like fairness, bias evaluation, or transparency into your engineering practice?
A: For me, data quality is where it all starts. In projects like RMGDIVE’s KYC system, we curated and validated onboarding data to avoid bias and ensure fairness. Fairness and bias are benchmarked and evaluated continuously, and we integrate monitoring mechanisms to track drift or unexpected behavior in production models, keeping them reliable over time. For generative AI components, prompt engineering and validation help reduce unintended outputs. In projects like AWAN Afrika, the plant disease detection and crop recommendation models were functional prototypes, so monitoring focused on testing accuracy, usability, and fairness in representative scenarios rather than full-scale production monitoring. This allowed us to validate assumptions and gather early feedback before scaling. In practice, ethical AI for me combines rigorous data practices, benchmarking, evaluation, and careful prototype testing, ensuring systems are reliable, fair, and impactful even in early stages.
Editorial commentary: From the very start, ethics is embedded in the system. Bias is measured before a line of code is written. Monitoring is built in before deployment. Data quality comes first, not last. Maximillian’s approach refuses the common shortcut of treating ethics as a compliance checkbox applied after the system is finished, a step too often deferred in both published frameworks and real-world practice.
In African contexts, the stakes of getting this wrong are immediate. The difference between a system that performs well on averages and one that serves all users equitably is where the real ethical work happens and it is work that requires constant attention, not a one-time audit.
Maximillian’s focus on monitoring for drift, validating prompts in generative AI, and testing prototypes for fairness alongside accuracy shows that ethical AI is not a feature to ship, but a practice to maintain.
Emerging Technologies & Africa's Future
Q: You regularly engage with emerging technologies like edge AI, quantization, IoT integration, and generative AI. Which of these advancements excites you most for Africa’s future, and why?
A: All of these technologies are exciting, but edge AI stands out for Africa because it addresses one of our biggest constraints: connectivity and infrastructure, especially in remote areas, making solutions more inclusive. By running intelligent models locally on devices, we can deliver real-time insights even where internet or cloud access is limited whether that’s precision agriculture, health diagnostics, or smart energy systems.I’m also intrigued by generative AI, particularly for localized content creation, education, and automating routine tasks. Still, the most immediate impact comes from solutions that work reliably on the ground, close to the user. For me, combining edge AI with practical, context-driven generative AI opens the door to scalable, low-cost solutions that truly empower African communities and businesses.
Ultimately, MIP Intelligence aims to become a standardized intelligence layer for African businesses, helping them scale, optimize, and compete globally
Latitude59 x Kenya conference highlighted how strong ecosystems are built where bold ideas meet the right conversations.
Editorial commentary: Edge AI may not get the same headlines as large language models, but its significance for Africa is immediate and practical. While debates often focus on what models know or which languages they support, the deeper question is where AI actually runs. In much of rural and peri-urban Africa, unreliable connectivity limits usefulness more than model sophistication.
By running intelligence on the device itself, edge AI bypasses that constraint, delivering real-time results even without stable networks. Combined with context-aware generative AI (local languages, simplified interfaces, and solutions rooted in everyday practices), it creates systems that work within existing infrastructure, not wait for ideal conditions.
Mentorship & Next Generation
Q: For young AI engineers across Africa who want to build meaningful solutions, not just demos, what guidance would you offer about mastery, resilience, and building in community rather than in isolation?
A: 1. Mastery through practice: Challenge yourself to get better at your craft every day. Don’t just follow tutorials experiment, break things, and understand why solutions work the way they do.
2. Solve real problems: Invest time in understanding the problem deeply before writing a single line of code. Meaningful AI solutions come from insights about people, context, and constraints not from clever demos.
3. Build in community: Don’t isolate yourself. Engage with peers, share learnings, and seek feedback. Communities accelerate growth, expose you to diverse perspectives, and help you spot real opportunities for impact.
Ultimately, meaningful solutions are the product of skill, curiosity, and collaboration, combined with a relentless focus on the problems you’re trying to solve.
Editorial commentary: The guidance may sound simple but its power lies in the order and emphasis. In African AI, credentials are often secondary; hands-on experimentation is how real competence develops. Deeply understanding the problem before writing a line of code challenges the “demo culture” that favors showy results over practical impact. Community accelerates learning through diverse perspectives, rapid feedback, and collaborative problem-solving. Maximillian’s principles form a philosophy of building AI differently where success comes not from different tools, but from a different relationship with the work itself.
Closing Remarks
The thread running through this conversation is a rare kind of discipline: starting with what is real, designing for constraints before they become crises, and measuring success by what works in actual conditions, not ideal ones. This approach is rare in an industry that often prizes bold ambition over careful execution, and it carries exceptional value for African AI, where delivering meaningful impact remains a persistent challenge.
What is being built is a methodology where each system creates tangible value, each community initiative reshapes how new builders approach their craft, and every platform under development provides infrastructure others can leverage. The philosophy guiding this work establishes a standard for what African AI can achieve at its best.
None of this is guaranteed. It requires conviction: the same conviction that drove Maximillian to pursue AI when adoption was minimal and skepticism was high. That conviction has only grown sharper, and on a continent whose AI future is still being written, it is precisely this kind of deliberate authorship that will define it.
TAIS is currently self-funded and intentionally independent. Your support helps me keep spotlighting African innovators and building a stronger ecosystem.
