End-to-end analysis with inspectable evidence.
The loan-risk and market-opportunity projects show the full path from source data and definitions to findings, recommendations and downloadable project evidence.
Data Projects · Analytics · Business Intelligence
Data projects built from source to decision.
My data portfolio brings together end-to-end analytics case studies and applied evaluation work across portfolio risk, market opportunity, search quality and AI outputs. In each project, I focus on the question, the evidence, the method and the decision the analysis needs to support.
Project range
Different projects. The same analytical discipline.
My work ranges from complete public case studies to multilingual search and AI evaluation for international platforms. Across both, I define the problem, test the evidence, document the reasoning and make the result clear enough to review and use.
Search and AI quality work across languages and markets.
Projects for TELUS Digital, Lionbridge, Appen, Leapforce and Outlier involved relevance, intent, factual accuracy, language quality and guideline-based judgement.
Clear logic from the source to the conclusion.
I separate observed signals from assumptions, explain the limits of the available data and structure outputs so the next decision is easier to make.
Public case studies
Real questions.Traceable analytical decisions.
These studies show how I move from an open business question to validated data, a reusable analytical layer and a recommendation or monitoring system that can be reviewed in detail.
Portfolio risk analytics
BigQuery and Looker case study
View case study →
A $3.08B portfolio needs connected risk controls.
I validated 270,299 loan records, created a reusable analytical layer in BigQuery and designed a Looker monitoring experience that connects an executive exposure alert to the status, geography, purpose and vintage behind it.
- $3.08B
- outstanding exposure
- 41.81%
- top-five state share
- 9.31%
- highest purpose loss rate
- Question
- How can a portfolio reviewer move from one aggregate threshold to the specific concentrations and segments that explain the risk?
- Analytical work
- Schema inspection, grain validation, risk-flag definitions, reusable marts, concentration analysis and an interactive monitoring layer.
- Decision output
- A framework that links total exposure to geography, loan status, purpose and origination cohort instead of treating the headline figure as a diagnosis.
- Scope
- A training dataset and an illustrative internal monitoring level. The project is not presented as a live credit model or regulatory calculation.
- SQL
- BigQuery
- Looker Enterprise
- Data quality
- Data modelling
Market opportunity analytics
International SEO case study
View case study →
The largest search market was not the strongest first move.
I cleaned and compared travel-search demand across Spain, France and the United Kingdom, then separated branded volume from accessible non-brand opportunity so the recommendation reflected the market a new entrant could realistically address.
- 2.15M
- monthly searches analysed
- 3
- markets compared
- 69.7%
- France non-brand share
- Question
- Which market offers the strongest accessible opportunity for an initial travel launch rather than simply the largest total search volume?
- Analytical work
- Data cleaning, keyword standardisation, branded and non-brand segmentation, competitor review and cross-market comparison.
- Decision output
- France emerged as the strongest first move, Spain as the second priority and the United Kingdom as a more selective opportunity.
- Scope
- A directional market-entry analysis based on the available keyword and competitor samples, with production research identified as the next step.
- Excel
- Data cleaning
- Market segmentation
- SEO analytics
- Competitive analysis
Search and AI evaluation
Multilingual projects built aroundrelevance, accuracy and judgement.
For TELUS Digital, Lionbridge, Appen, Leapforce and Outlier, I worked with detailed guidelines to assess search results, digital content and AI outputs across languages and markets.
Relevance, intent and usefulness across markets.
I evaluated search results and digital content against detailed guidelines, considering the query, user intent, language, location, result quality and whether the output actually answered the need.
- Contribution
- Guideline-based relevance assessment, intent interpretation, language-quality review and consistent documentation of edge cases.
- Context
- Multilingual datasets and search-quality projects where the same result can be appropriate in one market and misleading in another.
- Evidence
- The work strengthened a repeatable practice of separating observable signals from assumptions and explaining judgement clearly.
- TELUS Digital
- Lionbridge
- Appen
- Leapforce
Accuracy, reasoning and factual review under guidelines.
I reviewed generated or structured outputs for usefulness, language quality, factual consistency and adherence to task requirements, applying the same editorial discipline used in high-stakes publishing and quality assurance.
- Contribution
- Response evaluation, fact-checking, quality comparison, error identification and written justification.
- Context
- Tasks where a fluent answer is not enough: claims, reasoning, localisation and compliance with the requested format all affect quality.
- Project value
- The work combined fact-checking, language review and written justification to distinguish fluent answers from reliable, task-compliant ones.
- Outlier
- Multilingual datasets
- Quality assurance
- Fact-checking
Together, these projects strengthened my ability to apply complex guidelines consistently, explain judgement and detect quality issues across multilingual data.
How I structure the work
From source review to a decision someone can use.
Across the public case studies, I keep the analytical path visible: what the data contains, how the logic was built, what the result means and what should happen next.
The analytical starting point is explicit.
I document the dataset, the grain, the definitions used and the assumptions that shape the analysis.
Keeps every later result grounded in a clear foundation.The logic is organised beyond one chart.
Cleaning steps, calculations and analytical layers are structured so the work can be reviewed or extended.
Supports repeatability instead of a one-off answer.The finding connects to an action or choice.
The final presentation distinguishes what happened, why it matters and what should be investigated or prioritised next.
Keeps business context inside the analytical result.The conclusion stays useful and proportionate.
I explain what the available data supports, where the current analysis stops and how the project could be extended.
Shows the most useful next question instead of overstating the result.Contact
Need a clearer answer from complex data?
Tell me what needs to be understood, compared or monitored. I will bring the analytical structure, business context and communication needed to make the result usable.