Fleet risk overview
Risk distribution for predictive maintenance
The 'Financial Breakdown' (Waterfall)
demand pattern
Demand driver correlation
Route optimisation
Demand forecasting
Fleet-wide anomaly
Predictive maintenance
Time series decomposition
The 'Efficiency Frontier' (Scatter)

PROJECT UGUNJA

Inspiration

Our inspiration came directly from the hackathon sponsor, Greenwells Energies. We were struck by the story of its founder, Dr. David Onguka, who began in 2011 with a single station in the town of Ugunja.

He built an empire on a foundation of community trust. Our project is a tribute to that legacy.

We wanted to answer the question: "How do you protect that original 'Ugunja spirit' of trust and efficiency when you scale to a national level?" Our answer was to build an intelligence engine that tackles the modern challenges of trust, safety, and logistics.

What it does

Ugunja is an AI-powered logistics and authentication platform designed to solve the four core problems in the LPG industry:

Inefficient Deliveries: We solve the problem of high fuel costs and slow deliveries using an AI-powered Route Optimization model that creates the most efficient delivery paths for the entire fleet.

Unexpected Breakdowns: We solve the problem of fleet downtime using Predictive Maintenance. Our ML model analyzes vehicle data to flag trucks at high risk of failure before they break down.

Unpredictable Demand: We solve the problem of stock-outs and over-stocking with our Demand Forecasting model, allowing Greenwells to move from a reactive to a predictive supply chain.

Counterfeit & Unsafe Cylinders: We solve the problem of trust and safety with our "Digital Trust Certificate." A customer can scan a cylinder's QR code to instantly verify its authenticity and safety history.

Our solution is an integrated "Intelligence Engine" that directly aligns with the hackathon's core themes of Fleet, Delivery & Logistics Intelligence, Better Customer Support, Better Trust, Reimagining the Consumer Experience, Smart Operations & Workflow Management, and Data-Informed Decision-Making.

How we built it

The Intelligence Engine (Core): Our main focus was building Ugunja's "intelligence engine." This was more than just an app; it was a suite of machine learning models built in Python using libraries like Scikit-learn. We designed models to handle our core problems, experimenting with Support Vector Machines (SVM) for predictive maintenance analysis and ensemble methods (like Random Forests) for demand forecasting.

Backend: We used Django REST Framework (Python), which allowed us to build a robust, secure API and provides a clear path for integrating our Scikit-learn and TensorFlow machine learning models.

Frontend: The admin dashboard, customer app, and driver app were built using React and Next.js for a high-speed, modern user experience.

Database: We used PostgreSQL with the PostGIS extension to handle complex and efficient geospatial queries for fleet tracking and route optimization.

Authentication: We secured our platform using JWT (JSON Web Tokens) for all three user roles.

Demo & Vision: We prototyped the "Digital Trust Certificate" as a high-fidelity React component to demonstrate the "Proof of Authenticity" concept, with the vision of deploying it as an NFT on a green blockchain like Polygon.

Challenges we ran into

Our biggest technical challenge was architecting the deployment of our machine learning models into a real-time, production-level system. It's one thing to build an SVM or ensemble model; it's another to create a scalable infrastructure that can serve live predictions for route optimization and predictive maintenance across an entire fleet.

This involves setting up data pipelines, managing model versioning, and ensuring low-latency API responses.

For the hackathon, we solved this by containerizing our model logic and building a robust Django REST Framework API to serve the predictions. This proves that our "intelligence engine" is not just a concept—it's an integrated system ready for a full-scale cloud deployment on a platform like Azure or Render.

Integrating the Web3 "Trust Certificate" was a conceptual challenge. We focused on building the user experience of the verification process first, as this is what proves the value to the customer and the business, while clearly articulating the (future) on-chain architecture.

Accomplishments that we're proud of

We are incredibly proud of successfully framing our project as an "Intelligence Engine," not just another logistics app. The ML-driven insights are the product.

We're proud of how we wove a compelling narrative (the Ugunja origin story) into the very DNA of our product. It guided our design and features, making them more purposeful.

We successfully built and demonstrated the end-to-end "wow" factor: a clear, convincing demo of the "Digital Trust Certificate," which directly addresses the sponsor's biggest risk.

What we learned

We learned that the most powerful tech solutions are the ones that solve a fundamental human problem—in this case, the need for trust and safety in a product people use every day.

We learned how to "tell the story" of data. An ML model is just code, but a dashboard that shows "you will save 15% on fuel" or "this truck will break down" is a powerful business tool.

We solidified our skills in full-stack development, learning how to connect a robust Python backend to a sleek React frontend and manage complex geospatial data.

What's next for Ugunja

Ugunja is 70% complete and built to be pilot-ready. The next steps are:

Deploy the Pilot: Partner with Greenwells to deploy Ugunja on a small, live fleet to begin collecting real-world data.

Train Live Models: Feed this live data into our ML pipelines to train and refine our Demand Forecasting and Predictive Maintenance models.

Mint the "Trust Certificate" NFTs: Fully build out the Web3 backend, minting a "Digital Trust Certificate" for every new cylinder on a low-cost, eco-friendly blockchain (e.g., Polygon).

Integrate Payments: Integrate M-Pesa to create a truly seamless, end-to-end digital experience for customers.