Projects

8 projects, starting with the work I'd lead with. The label on each says what it asked of the data: describe what happened, predict what will, recommend what to do about it, or test whether the answers can be trusted at all.

Applied AIIndividual project · Generative AI and LLMs, Imperial Business School

Confidently Wrong: Testing a World Cup Rules Chatbot

The version that answered the most questions correctly was also the one that invented an answer to every single question it should have refused.

Retrieval-Augmented GenerationLLM EvaluationPrompt EngineeringPythonAI Safety Testing

Read the project →

PrescriptiveCapstone project · Analytics in Business, Imperial Business School

Heat Smart Orkney: The Commercial Case for Demand Response

A residential demand-response scheme would absorb less than 1% of the curtailed wind energy it could technically address. It would still turn an annual profit of £151,000 at scale.

PythonFinancial ModellingSensitivity AnalysisData CleaningExecutive Reporting

Read the project →

PredictiveGroup project · Data Wrangling and Visualisation, Imperial Business School

Customer Segmentation & Heterogeneous Churn Dynamics

In most customer segments, ordering more predicts a higher chance of churning. In the highest-risk segment, it's the opposite: ordering more predicts they'll stay.

Cluster AnalysisLogistic RegressionPredictive ModelingRCustomer Segmentation

Read the project →

PredictiveGroup project · Healthcare & Medical Analytics, Imperial Business School

Psychological Distress & Physical Multimorbidity in England

Each one-point rise in psychological distress predicts a 9.7% increase in physical health conditions. That effect is statistically identical in the richest and poorest neighbourhoods.

PythonStatistical ModellingRegression AnalysisSurvey DataHypothesis Testing

Read the project →

DescriptiveGroup project · Text Analysis for Business, Imperial Business School

Do Listings Lie? How Airbnb Language Predicts Price

Premium words like “luxury” and “stunning” predict nightly price three times more strongly than generic positive sentiment.

Text AnalyticsRLASSO RegressionTopic ModellingExploratory Data Analysis

Read the project →

PrescriptiveIndividual project · Data Analytics Simulation, Imperial Business School

HBS Data Analytics Simulation: Brand Strategy & Market Forecasting

Blue was the weakest of four detergent brands, flat at 9-11% market share for four years. Repositioning it on data instead of price took market share to 14.1% and cumulative profit to $242.0M by Year 8.

Strategic Decision-MakingScenario AnalysisMarketing AnalyticsBrand StrategyData Storytelling

Read the project →

DescriptiveIndividual project · Self-directed SQL analysis

How This Product Is Actually Doing: A SQL Investigation

One user segment uses the product six times more than another. Both open their email at almost the same rate.

SQLWindow FunctionsData CleaningProduct AnalyticsExploratory Data Analysis

Read the project →

DescriptiveIndividual project · Self-directed SQL analysis

Do the Numbers Back the Legends? Thirteen Years of IPL in SQL

AB de Villiers leads the strike-rate table among the tournament's highest-volume scorers, leads it again under the highest-pressure overs of an innings, and has more Player of the Match awards than anyone in IPL history. That holds across 13 years of data, not one standout season.

SQLWindow FunctionsData StorytellingSports AnalyticsExploratory Data Analysis

Read the project →