Professional Case Studies
The projects below are drawn from my time as a Data Analyst at Mime, where I worked with schools, local authorities, and third-sector organisations across the UK. Each one was delivered to a real client, with real constraints, and built to be maintained independently after handover.
Governance Action Tracker - Automated Survey Workflow
Client: Education sector client (via Mime)
Type: Automation & Data Analysis
The problem: Governance reporting was a manual, time-intensive process. Every time a new survey was needed, responses had to be collected, aggregated, and processed by hand, and any updates to the governance framework meant starting over. It worked, but it was slow and fragile.
What I did: I completely redesigned and automated the backend workflow from scratch. The new system handles survey creation, response aggregation, data processing, and framework updates end-to-end, with automated analysis feeding directly into a custom dashboard with click-through detail views. Everything is documented so it can be run and maintained independently.
The result: A task that previously consumed significant analyst time now takes under 30 minutes to complete, end-to-end. The client has a reliable, repeatable system that they own and can operate themselves.
Skills: Excel, VBA, workflow automation, dashboard design, automated analysis, technical documentation
Survey Data Processing Toolkit
Client: Education sector clients (via Mime)
Type: Automation & Tooling
The problem: Survey data preparation was being done manually — a slow, error-prone process that created a bottleneck before any analysis could begin.
What I did: I designed and built a complete automated ETL and QA pipeline from scratch. This included Excel/VBA utilities for data cleaning, Web API integration to backfill missing data, and validation checks throughout. Crucially, I built it so that clients could run it themselves — with clear documentation and a user-friendly interface — rather than being dependent on me or my team for every update.
The result: A fully client-operable toolkit that eliminated manual data preparation, reduced errors, and freed up analyst time for actual analysis. Final outputs fed directly into a linked Tableau dashboard ready for handover.
Skills: Excel, VBA, Web APIs, ETL pipeline design, QA and validation, Tableau, technical documentation
Inclusion Index - Educational Trust Marketing Tool
Client: Multi-Academy Trust (via Mime)
Type: Analysis, Python, Data Product
The problem: A client wanted a way to measure inclusivity across schools and use that insight for marketing outreach — but no such metric existed.
What I did: I created an "Inclusion Index" by combining multiple public data sources into a single composite metric, designed from scratch. I then used Python, web scraping, and APIs to compile contact details for approximately 300 educational leaders for marketing outreach, and built a free public-facing interactive dashboard to showcase the findings.
The result: A complete data product — from bespoke metric design through to a live public dashboard — that gave the client both a credible insight tool and a marketing asset.
Skills: Python, web scraping, APIs, composite metric design, Tableau/dashboard building, strategic thinking
AI & Prompt Engineering Blog Series
A growing series on how to get the most out of modern LLMs — grounded in research and practical experience rather than hype.
-
Forget "Think step by step" - Here's How to Actually Improve LLM Accuracy: an evidence-based look at why Chain-of-Thought prompting has become ineffective on modern models, and what actually works instead
-
You're Using ChatGPT Wrong. Here's How to Prompt Like a Pro: advanced prompt engineering techniques including role-based prompting, task decomposition, ReAct, and few-shot prompting
-
Still Copy-Pasting into ChatGPT? Here's How to Turn Your Ideas into AI-Powered Apps: a practical guide to accessing LLMs via API in Python to build automation tools and custom data pipelines
-
Getting Started with AI: The Fundamentals of Prompting: a beginner's guide to how LLMs work and how to write prompts that actually get results
Algorithms, Mathematics & Data Science Blog Series
-
Solving the Travelling Salesman Problem Using a Genetic Algorithm: algorithmic design, Python, visualisation
-
Procedural Terrain Generation with Perlin Noise: mathematical algorithms, parameter exploration, 3D visualisation
-
Data Visualisation Best Practices: pre-attentive attributes, Gestalt principles, applied to real charts
-
Image Processing & Computer Vision: edge detection and K-means segmentation in Python
-
Graph Theory & Mathematical Proofs: Bridges of Königsberg, Stable Matching, Tower of Hanoi
