I'm Elangovan Murugan, a data scientist currently working in public-sector analytics at the Yorkshire & Humber Office for Data Analytics, and finishing an MSc in Data Science at the University of Sheffield — mapping connectivity, risk and behaviour, from my dissertation on the UK's rail network to NHS wards and airline cabins.
I'm an associate-level data scientist currently working in public-sector analytics at the Yorkshire & Humber Office for Data Analytics, and finishing an MSc in Data Science at the University of Sheffield. I like the full pipeline: pulling together messy public data, building it into models and dashboards, and explaining clearly what it means to someone who has to make a decision from it.
Outside pure modelling, I've spent a lot of time coordinating people under pressure — running international conferences for 700+ delegates, representing 50+ postgraduate students to faculty, and stewarding crowds at events like the British Grand Prix. That mix of technical depth and on-the-ground coordination is what I try to bring to a team.
Curiosity and structure — in the data, and in the room.
The dissertation is the flagship — everything else here is applied ML and analytics across healthcare, transport, aviation and research data.
Built a nationwide weighted rail network of 2,588 stations and 3,658 connections in NetworkX, then tested four different definitions of the UK's North–South boundary against it using graph-theoretic measures and Random Forest / SVM models. The geographic boundary definition separated the network with a 100% F1-score — a data-backed way to talk about regional transport inequality.
Analysed 5,990+ hourly OpenAQ/NHS air-quality readings, flagging 329 extreme PM2.5 events and showing Sheffield's annual mean PM2.5 ran 48% above WHO guidance — built as a composite visualisation grounded in Grammar of Graphics and WCAG accessibility principles.
Processed 55,000+ ICLR papers and OpenAlex metadata on Databricks serverless PySpark, building join, aggregation and window-function pipelines to surface trends across academic literature at scale.
Modelled 30-day hospital readmission risk on NHS patient data, reaching 85% accuracy, then wrapped it in a SHAP-explainability dashboard so clinicians could see why a patient was flagged, not just that they were.
Compared Decision Tree and Random Forest models on 5,196 airline passenger records, hitting 89.9% accuracy and 0.96 AUC, and identified the service-quality factors that mattered most to passengers.
Built and validated a loan-default classifier, cutting misclassification by 20%, then shipped it as a Flask API with MLflow tracking so the model could be versioned and deployed, not just prototyped.
Designed structured analytical pipelines in Python and Power BI to monitor funding distribution and surface data-quality issues across 100+ organisations. Improved stakeholder decision efficiency by 30% and delivered recommendations that directly informed 50+ community initiatives.
Represent 50+ postgraduate students in formal faculty committees, turning student feedback into structured, evidence-based submissions on assessment transparency and academic engagement.
Led end-to-end planning and delivery of two large-scale international academic conferences — coordinating multidisciplinary teams, speakers, sponsors and logistics under competing priorities and tight timelines.
Performed EDA on 10,000+ booking records to identify trends and behaviour patterns, then built demand-forecasting models that improved forecast accuracy by 18%.
Designed and deployed agent-based AI systems, automating 30% of manual decision workflows across stakeholder-facing processes.
Curiosity beyond the data too — trekking, travel and time spent on the ground at a Formula 1 weekend.
Part-time event steward with FGH Security at this year's British Grand Prix — crowd management and event operations across a full Formula 1 race weekend.
Regular trekking and hill-walking across the Peak District's routes, close to home in Sheffield.
Recent trips to Bath and Oxford — Roman history in one, collegiate architecture in the other.
Happy to talk about data science, machine learning, analytics or research opportunities — or just compare notes on rail networks and race weekends.