JKSH AI wasn’t just built to show off technical skill—it was born from necessity. It’s a direct response to Africa’s layered infrastructural challenges, where limited access to stable electricity, expensive data, intermittent connectivity, and language diversity are all part of the day-to-day reality. Many global systems assume high-speed internet, modern devices, and English fluency, but in our context, these assumptions break down. JKSH AI is purposefully engineered to do the opposite: to thrive despite these constraints.
At the core of JKSH AI is an intentional design philosophy rooted in resource-constrained computing. This became our driving theme as part of the Africa Deep Tech Challenge 2025. We focused not only on what the system does—offering intelligent financial advice and automation—but how it does it: through extreme efficiency, resilience, and accessibility.
To operate within Africa’s bandwidth limitations, JKSH AI minimizes unnecessary data exchange. The architecture emphasizes local caching and lightweight data payloads. Where internet access is weak or unreliable, it defaults to offline-first behaviors and queues up actions for when connectivity is restored. It’s not just cloud-smart—it’s edge-aware.
Because smartphones aren’t universal and apps with heavy UI are not always practical, we chose WhatsApp as our user interface. It’s familiar, cheap, and already widely adopted. Users can talk to JKSH AI in plain language, in a chat format they already trust. From checking savings insights to scheduling mobile top-ups, all it takes is a simple message—no apps, no downloads, no friction.
Additionally, the system is designed to scale with existing informal structures. Whether someone is a street vendor in Nakuru or a boda boda rider in Kampala, JKSH AI aims to speak to them in context—through language, tone, and timing that resonates.
We chose Rust for its performance and compiled efficiency, allowing us to build a backend that consumes very little memory and CPU. PostgreSQL gave us robust data storage, and Docker allowed quick deployment even on modest infrastructure. The system was further optimized to function seamlessly over 2G and unstable networks—without losing functionality.
WhatsApp was selected as our primary interface due to its wide adoption and low data footprint. Instead of building a new app that requires installation and updates, we integrated with the WhatsApp Cloud API, giving users instant access to JKSH AI via a simple chat number.
The most important part of the journey was engaging real users—individuals living in rural and peri-urban Kenya, where digital tools must fit into complex social and economic realities. These are people who may not speak English as a first language, often rely on shared or low-end smartphones, and purchase data in small, incremental bundles that are rationed carefully. In these environments, technology cannot afford to be abstract—it must be practical, culturally grounded, and emotionally intuitive.
We didn’t just translate the interface into Swahili. We co-designed it with the people we serve. We listened to how they express financial goals, how they describe trust, urgency, and savings. We adapted JKSH AI to use common local expressions, reflect regional financial behavior, and respond in a tone that feels human—not robotic. Users aren’t greeted with jargon or generic answers. Instead, they receive responses that are warm, direct, and relatable.
Many users we worked with had never used AI systems before. Some were hesitant at first, unsure if this 'chatbot' could really understand them. But over time, as they asked questions, received helpful insights, and saw real-time actions like airtime top-ups or loan reminders work correctly, they began to trust the system—not just as a tool, but as a companion.
This phase of field testing proved invaluable. It exposed assumptions we hadn’t considered, and it helped us refine not only how JKSH AI speaks, but how it listens. In the end, it reminded us that the best technology doesn’t just function—it connects.
This project is about more than just software. It's about designing technology that listens first, adapts second, and scales last. It’s about building for the present, not some imagined future. JKSH AI shows what’s possible when you embrace Africa’s infrastructural reality instead of trying to engineer around it.
JKSH AI is built for the people who need it most—and that means embracing resource constraints not as a limitation, but as a design principle.