Announcing the winners of “Hack Together: The Microsoft Data + AI Kenya Hack”
We are excited to announce the winners of the winners of Hack Together: The Microsoft Data + AI Kenya Hack!
About the Hackathon
The HackTogether was an exciting opportunity to bring bold ideas to life by building Data & AI solutions using Microsoft Fabric and Azure AI Services. Organized exclusively for participants from Kenya, the hackathon launched on March 12th during the AI Tour Stop in Nairobi and ran for three weeks, ending on April 11th with 688 participants registering.
Throughout the event, we hosted weekly livestream sessions designed to help the participants get started with Microsoft Fabric and Azure AI. Some of the sessions included A tidy introduction to Microsoft Fabric, Building Scalable Data Solutions with Microsoft Fabric’s Data Warehouse, Building high scale RAG Applications with the Eventhouse in Microsoft Fabric and Building AI Agents using Azure AI Agent Service.
To ensure that all participants had access to the latest Azure Open AI Models, we provided an Azure Open AI Proxy Service, which allowed participants to use the Azure Open AI Service without the need for an Azure subscription. Over the course of the Hackathon, hackers made over 95,000 requests to the Azure Open AI proxy!
The Winners
We received 32 project submissions, and our judges had the difficult task of selecting the top three projects. The winners were announced on April 25th during the official Closing & Awards Ceremony at the Microsoft Africa Development Center in Nairobi.
The judging process considered the following criteria: Alignment with the hackathon goal of building and deploying a functional real-world AI solution using Microsoft Fabric, Technology (Fabric + Azure AI), Innovation/Impact, Documentation and video Presentation. We were impressed by the quality of the project submitted, considering that over 95% of the participants were interacting with Microsoft Fabric for the first time.
Each of the submitted projects was unique and showcased the creativity and innovation of the participants. The projects demonstrated that Microsoft Fabric is the Analytics platform for the era of AI, whether you are working with real-time data, batch data, or a combination of both.
Even though we can only have the top three winners, we would like to acknowledge with an honorable mention to some other projects that stood out during the judging process.
🏆 Overall Winner: RAG-Powered Virtual Assistant for the Higher Education Fund
The winning project was a virtual assistant that helps users understand Kenya’s new Higher Education Funding Model. The team loaded documents into Microsoft Fabric Lakehouse, split them into smaller chunks, and created vector embeddings. They stored these embeddings in an Eventhouse, which they used as a vector store. When a user asks a question, the assistant turns the question into an embedding and finds the most relevant parts from the Eventhouse. It then sends this to GPT-4o through Azure OpenAI to get a clear, helpful answer.
It was great to see the participants pick up Eventhouse as a vector store during the livestreams and then turn that knowledge into a polished, prize‑winning solution.
Project Repository URL
Project video
🥈 1st Runner Up: Anomaly Detection in Financial Institution Data
This project is a lightweight, real-time system that helps financial institutions spot unusual transactions as they happen. It doesn’t try to predict fraud but rather detects and flags it fast. The team built the solution using Microsoft Fabric and once the system receives realtime transaction data, it runs a machine learning model, to detect anomalies using the data science workload in Fabric. These anomalies are then sent to Azure OpenAI’s GPT‑4o‑mini, which ranks them as high, medium, or low risk. It also adds a short explanation for each one.
High-risk anomalies are saved in files and stored in Azure Blob Storage. Power Automate has a blob storage trigger, and when a new file is added or changed, it sends out alerts. These can go via email or Microsoft Teams, depending on what the risk team prefers.
Project Repository URL
https://github.com/SamuelNw/Anomaly_Detection
Project video
🥉 2nd Runner Up: Agentic-Powered Disease Surveillance and Outbreak Forecasting in Kenya
This project aims to build a real-time, AI-powered disease surveillance and outbreak forecasting solution tailored to Kenya’s healthcare landscape.
The solution ingests raw datasets through a medallion architecture using Fabric’s Dataflow Gen2. It stores and cleans the data in the Lakehouse. The cleaned data is enriched in the Eventhouse for AI-powered retrieval. Built‑in ML models in Fabric forecast outbreak trends and Azure OpenAI adds natural‑language chat so users can talk to the system.
A web UI lets users ask questions in plain English and get answers, while Power BI dashboards show the results.
Project Repository URL
https://github.com/Malvine254/Disease_Surveillance_Forecasting
Project video
💝 Honorable Mentions
We received many high-quality submissions and encourage you to check out all the hackathon projects and see if there are any that interest you. Each of these submissions has provided value not just for the hackers but for anyone interested in looking into how their problems were solved using Microsoft Fabric workloads along with Azure AI Service.
🪙 Data-driven farming innovation with Microsoft Fabric and AI
This project is a data-driven solution built using Microsoft Fabric and Azure OpenAI to forecast avocado sales and production trends in Kenya. It brings together data from multiple sources: including historical sales records, weather data, and global market forecasts. This is done through pipelines, dataflows, notebooks, and event streams. The data is organized using a medallion architecture, moving through raw, cleaned, and enriched layers to support structured analysis and modeling.
The solution applies machine learning models, such as linear regression, to predict future trends from 2025 to 2030. Predictions are stored in the Fabric Lakehouse and visualized through Power BI. Azure OpenAI is used to generate natural language summaries and business insights from the prediction output, making it easier to communicate trends. The entire pipeline is modular, scalable, and built for seamless integration with modern data workflows.
Project Repository URL
https://github.com/DaymondMartin/Avocado-Sales-Predictions
Project video
🪙 AI-Powered Dermatology Assistant
This project is an AI-powered dermatology assistant that helps users get early insights about skin conditions. It uses a deep learning model to classify skin diseases from images and combines this with Azure OpenAI to give easy-to-understand health recommendations. The system is built with Django for the backend and React.js for the frontend, creating a smooth experience for users to upload photos, ask questions, and get quick responses.
Behind the scenes, the data flows through Microsoft Fabric, where it’s collected, cleaned, and organized for analysis. Predictions and trends are visualized in Power BI dashboards to make the information clear and actionable. Everything is securely hosted on Azure, making the platform reliable, scalable.
Project Repository URL
https://github.com/COMFORTINE-SIWENDE/AI-Powered-Dermatology-Assistant
Project video
🪙 GenoHos-Genomic Analysis
GenoHos is an AI-powered healthcare analytics platform designed to transform multi-omics data into actionable clinical insights for personalized oncology care. It addresses the challenge of fragmented genomic, proteomic, and clinical data by integrating siloed systems (EHRs, sequencing machines, research databases) into a unified AI-driven platform. The app enables real-time genomic analysis, molecular phenotyping, and a Retrieval-Augmented Generation (RAG) chat interface to empower clinicians with precision medicine tools.
Project Repository URL
https://github.com/danielmuthama23/GenoHos-Microsoft_HackAI-Genomic_Analysis
Project video
🪙 Data-Driven Decisions with AnalytiQ
The team built an AI-powered feedback analysis platform that helps businesses make sense of large volumes of unstructured customer input; like reviews, surveys, and reports. Built with Azure OpenAI and a serverless architecture, the system enhances each analysis by retrieving relevant past feedback, allowing for deeper sentiment understanding and more accurate issue detection.
Data flows through Azure Blob Storage, Azure Functions, and an intelligent document parser that supports multiple formats. Results are stored in Azure SQL and visualized through Power BI Fabric, with real-time dashboards that include tenant-level filtering and row-level security. The platform is fully integrated with Microsoft Fabric’s ecosystem, using Direct Lake mode, semantic models, and embedded analytics to ensure that insights are not only powerful but also accessible and secure for every organization.
Project Repository URL
https://github.com/victorbash400/AnalitiqProject
Project video
Congratulations 💜
Congratulations and thank you to all who joined us for the “HackTogether: The Microsoft Data + AI Kenya Hack”! Whether you won a prize or not, you’re all winners to us 🎉! You made the deliberate effort to learn a new skill and put in the work and effort towards it, kudos to you!
A special thank you to our judges for their help with evaluating the projects and selecting the top winning entries: Lee Stott, Paul DeCarlo, Jasmine Greenaway, Christopher Maneu, Someleze Diko and Marc Baiza.
We’d love to hear what you think in the comments. Let us know if there is a particular solution that stands out to you how you think it could provide value to end-users!
