A summary of what I've built.
2026
-
EPC Assessor App
Survey list, field entry with CV / AI-verify controls, and report generation on mobile. Click to expand. Every home in the UK that is sold or rented needs an Energy Performance Certificate (EPC), which grades how energy-efficient the property is. Producing one means a surveyor visiting the property and manually recording dozens of details about the roof, walls, windows, boiler, and more, which takes over three hours. This prototype web app helps the surveyor by automating 21 fields while doing the survey, using just the phone camera. The app uses computer vision to auto-classify six property features (roof, walls, windows, solar, property type, subtype), OCR to read the boiler's data plate and identify the home's heating system, and the Open Data API to fill in property details. Trained three different image-classification architectures on a dataset of 5,000+ images and compared two OCR engines to find the best-performing one for boiler data plates. The prototype cuts down manual data entry and saves up to an hour per survey.
-
Superstore Sales Dashboard
Overview, Products, and Customers. Click to expand. A three-page Power BI dashboard on the public Superstore sample dataset (5,009 orders, 2011 to 2014, $2.3M in sales). Overview tracks sales, profit, margin, and orders with a monthly trend and breakdowns by category, region, and segment. Products flags sub-categories running at a loss and ranks items by both sales and profit. Customers identifies the 155 loss-making accounts out of 793, and plots the sales-vs-profit spread. Filters at the top of each page cascade through every visual.
- Understanding the data and its connections.
- Breaking it down for insights.
- Presenting only what needs to be measured.
2025
-
Operations Dashboard for an Energy Assessment Company
Confidential data
An energy assessment company was tracking sales and payments through invoices, WhatsApp chats, and memory, with no month-on-month record of revenue. Built a Power BI dashboard focused on what actually mattered: monthly sales, per-project profitability, pending payments by client, and where projects were getting stuck in the pipeline. The dashboard made things visible that weren't before: sales performance across six different channels; profitability broken down across more than four cost heads; loss-making and un-submitted projects; and which clients owed money. Set up the full pipeline from Excel through Microsoft Fabric to a Power BI mobile app, and later shifted the whole thing to Google Looker Studio.
2023
-
Sales Dashboard for an Educational Institute
Confidential data
An educational institute was tracking sales in Excel by hand, using only totals. Built a Power BI dashboard on top of the same data, focused on what actually mattered: monthly sales, top-performing teachers, courses, and subjects, and pending instalment payments. The dashboard made things visible that weren't before: month-on-month sales, so improvement was actually measurable; seasonal trends across the year; top-earning courses and subjects; and unpaid instalments that were being missed.