Works

Things I've built

Data pipelines, models, dashboards and automations. Filter by the kind of work you care about.

Research · Machine learning

Rice disease detection, from field to phone

An IoT system that reads soil moisture, NPK, temperature and humidity, then pairs that with image-based disease detection and a rule-based recommendation engine, so a farmer gets advice and not just a reading.

I fine-tuned a MobileNetV2 model on 22,688 images across 15 rice diseases. It reached 96.42% accuracy and runs real-time inference.

  • Python
  • MobileNetV2
  • Transfer learning
  • IoT sensors
Read the paper ↗

Data · Machine learning

Loan default risk model

A full pipeline in Python: cleaning, feature engineering, then Scikit-learn and XGBoost models to predict who is likely to default. Power BI dashboards show the drivers, such as credit score and debt-to-income ratio, in plain view.

  • Pandas
  • XGBoost
  • Power BI

Data · Government

National education data frameworks

For UBEC, I digitised and quality-checked reporting frameworks, built ODK collection tools with strict validation so bad entries are caught at the source, and made dashboards that government stakeholders can use.

  • ODK
  • Data validation
  • Dashboards

Automation

AI-assisted client onboarding

An automated workflow using Chatbase and Zapier that removed manual data entry from onboarding, plus Zoho CRM set up with lead scoring and pipeline management.

  • Chatbase
  • Zapier
  • Zoho CRM

Research · Hardware

Automatic railway crossing gate

A sensor-based gate system for car crossings, from the literature review to prototype testing. It performed with 100% functional accuracy in controlled tests.

  • Embedded systems
  • Sensors
Read the paper ↗

AI internship

Web scraper and face recognition

During my internship at NCAIR, I wrote a Python script for automated data extraction from websites and worked with the team on a facial recognition system using computer vision.

  • Python
  • Computer vision