Machine learning · People analyticsPredicting AI Job Salaries
I used a dataset of about 90,000 AI and data-job records to compare Linear Regression, Random Forest and XGBoost. My aim was not only to predict salary, but to understand which factors were carrying the most weight.
RXGBoost R² 0.94Random ForestFeature importance
What I learned: XGBoost performed best, while experience, bonus and country were the strongest salary drivers. The project strengthened my interest in compensation and people analytics because it showed how technical modelling can support better reward decisions without replacing human judgement.
Forecasting · OptimisationRetail Sales Forecasting & Profit Optimisation
As part of a ten-person analytics team, I helped build a multi-model retail analysis using more than 9,000 transactions. We examined demand, discounting, profitability, customer segments and transaction risk.
SARIMAMAPE 18.97%MASE 0.43K-Means
What I learned: SARIMA captured seasonal demand better than ETS, while regression showed that discounting was a major margin risk. We turned the findings into a 20% safety-stock rule, discount controls and targeted customer actions.
Customer analytics · ClassificationTelecom Customer Churn
I compared Logistic Regression, Decision Tree and Random Forest models on 7,043 telecom customers to identify customers more likely to leave and the factors associated with that risk.
PythonAUC 0.84Class weightingCross-validation
What I learned: Random Forest had the highest raw accuracy, but Logistic Regression gave the stronger balance of discrimination and interpretability. Contract type, tenure and monthly charges became the basis for targeted retention recommendations.
Risk analytics · ProfitabilityMedical Insurance Profitability
Working within a large team, I analysed a 100,000-record medical-insurance dataset to understand why the portfolio was losing money and which customer behaviours were most closely associated with medical cost.
R100,000 recordsRisk segmentationBreak-even modelling
What I learned: healthcare utilisation was more informative than demographic variables such as age or BMI. The analysis reframed the problem from broad demographic pricing to utilisation-based risk and cost management.
SPSS · Statistical inferenceStudent Success Factors
In Team Genesis, I worked on an SPSS analysis of 6,607 students. My main analytical contribution was the multiple regression model used to assess the combined influence of attendance, study hours, previous scores, tutoring and sleep on exam performance.
SPSSR² 0.597RegressionANOVA & t-tests
What I learned: attendance and hours studied were the strongest predictors, while some statistically significant relationships had very small practical effects. That distinction between statistical significance and useful business meaning became one of the most important lessons from the project.
HR analytics · RWorkforce Analytics & Employee Retention
I used R to explore roughly 100,000 employee records covering demographics, department, performance, satisfaction, remote work and resignation. I expected satisfaction and flexible work to explain turnover, but the data did not support that assumption.
RStudiotidyverseggplot2~100k records
What I learned: a useful analysis does not need to confirm the expected answer. Satisfaction and remote-work frequency were weak explanations of resignation in this dataset, so I recommended redirecting attention to factors such as progression, workload, compensation and role-specific pressures.
Research design · Labour marketGlobal AI Salary Research Design
I designed a cross-national study of salary differences in the global AI labour market, combining human-capital, firm, labour-market and regional factors rather than studying them separately.
OLSBartik IVRobust diagnostics90,000 observations
What I learned: good analytics starts before modelling. I had to define the research problem, identify gaps in existing evidence, justify the chosen design and confront endogeneity rather than simply run a regression because the data was available.
Strategy · Financial analysisVoltex Motors Business Simulation
I analysed four rounds of competitive automotive decision-making, connecting production, pricing, finance, market share and team choices to performance. The company moved from a loss in Round 1 to strong profitability in Round 4.
£549.88m R4 profit100% sell-throughScenario modelling+123% shareholder funds
What I learned: meeting internal objectives is not enough if competitors move faster. The company improved dramatically but still ranked fifth of six on post-tax profit, which taught me to compare strategy against the market, not only against my own previous performance.
Live consultancy · Team leadGrasmere Academy Marketing Strategy
As Team Lead and Senior Consultant for Metrics Mind Consulting, I coordinated a live client project for Grasmere Academy. We investigated why a school rated Good by Ofsted remained well below capacity while neighbouring schools were full.
SWOT & TOWSStakeholder mappingCompetitor auditCosted strategy
What I learned: the problem was not the quality of the school; it was visibility, positioning and conversion. I helped bring together research, competitor evidence, client meetings and financial framing into Bronze, Silver and Gold recommendations beginning at £150.
Leadership · Professional practiceLeadership Capability & Self-Analysis
I used a range of psychometric and reflective tools to examine how I lead, communicate, learn and make decisions. Rather than presenting only strengths, I used the weaker results to build a development plan.
Emotional intelligenceBelbinBig FiveReflective practice
What I learned: my strongest patterns were self-awareness, conscientiousness, service orientation and respect for diversity, while initiative, visibility and assertive influence were clearer development areas. I use that insight to guide how I work in teams and client-facing environments.
QMS · ComplianceQuality Management & ISO/IEC 17020
My quality-management experience taught me to see compliance as a working system, not a folder of documents. I have worked with the relationship between standards, policies, SOPs, work instructions, records, training, audits and corrective action.
ISO/IEC 17020SOPsWork instructionsAudit readiness
What I learned: ISO/IEC 17020 sets the requirements, while the QMS is the practical system that proves those requirements are being met consistently. This perspective continues to shape how I approach data quality, traceability, governance and administrative control.
HR · Administration · Data handlingPeople Operations & Evidence-Led Administration
Before specialising in analytics, I worked directly with HR and administrative processes: recruitment and staffing records, performance information, compensation documentation, leave, welfare, reporting, document control and audit support.
Workforce reportingRecordsPerformance dataDocument control
What I learned: analytics is only as reliable as the process that creates the data. My administrative background gives me an operational understanding of where workforce information comes from, why accuracy matters, and how seemingly small record-keeping errors can affect decisions later.
Digital innovation · Product thinkingRD NovaSphere & ScholarSpace
I designed and developed the RD NovaSphere website you are using now. I wanted it to be more than a static portfolio, so I built it around an innovation ecosystem where projects, professional portfolios and experimental digital ideas can live in one coherent space.
Website designFront-end developmentScholarSpacePrivacy-conscious publishingInteractive UX
What I learned: building the site forced me to think like both a user and a product owner: how people discover work, how a portfolio tells a story, how contributors retain control over what is public, and how design can make complex information feel simple. ScholarSpace grew from that idea — curated, searchable professional evidence rather than a document dump.