Oluwajuwon Adediji
Hire / Contact
Portfolio / Proof of Work

Projects that show how I think, analyse and communicate.

Every case study is framed around a problem, method, evidence status, limitation and decision implication. Synthetic work is explicitly labelled.

Review projectsEvidence hubGitHub
Recruiter snapshot

Proof signals

50K
synthetic borrowers in credit-risk modelling
3K
orders in SQL + BI work
240
AI benchmark cases evaluated
160
documents benchmarked

These figures describe portfolio experiments and benchmark work, not client scale or production business outcomes.

Case studies

Current evidence

Start with the business question, then inspect the method and primary artifact.

Data AnalyticsSQLPythonBIData Quality
Implemented / Public repository

AI-Powered Retail Sales Diagnostic

Question: Why can revenue decline while units sold increase?

Pipeline

Profile → clean → validate → KPI tables → decomposition → anomaly detection → dashboard/report.

Artifacts

Generated report, dashboard, monthly KPIs, revenue bridge, hypothesis effects and validation log.

Repository

Synthetic observational case with documented causal limits.

Decision layer: the project separates descriptive revenue drivers from causal claims and explicitly identifies where controlled testing would be required.

Limitation: findings describe patterns in generated data; they do not establish causation or production impact.

Inspect GitHub Open dashboard Read report
Business IntelligenceSQLSQLiteKPI Analysis
Portfolio evidence

SQL + BI Revenue & Order Analytics

Question: How should revenue, customer segments and regional performance be diagnosed from a relational order dataset?

Scale

3,000 orders modelled.

Methods

Relational modelling, SQL KPI queries, segmentation and regional analysis.

Decision use

Revenue and performance diagnosis for business reporting.

Next evidence layer: dashboard screenshots and an expanded case-study narrative should be attached as primary visual artifacts.

See evidence standard
EconomicsForecastingScenario AnalysisValidation
Portfolio evidence

Economic / Financial Forecasting & Scenario Analysis

Question: How can a forecasting workflow be validated and translated into decision-support scenarios rather than treated as a point estimate?

History

60-month historical/backtest window.

Validation

12-month holdout validation.

Decision layer

Scenario analysis and interpretation.

Limitation: model performance is context- and data-dependent and should not be interpreted as a guarantee of future forecasting accuracy.

See evidence standard
AI AutomationDocument IntelligenceValidation
Portfolio evidence

AI Document Intelligence Workflow

Question: Can document classification and extraction be structured as a repeatable pipeline with explicit validation rather than unverified model output?

Scale

160 documents benchmarked.

Workflow

Classification → extraction → validation → structured output.

Quality principle

Separate AI generation from QA and validation.

See evidence standard
AI EvaluationBenchmarkingQuality Metrics
Portfolio evidence

AI Output Evaluation Benchmark

Question: How can AI workflow quality be compared with deterministic measures instead of subjective impressions alone?

Scale

240 benchmark cases.

Metrics

Semantic similarity, keyword coverage and instruction adherence.

Use

Compare workflow quality and identify failure modes.

See evidence standard
How to evaluate the portfolio

Trace each claim to an artifact.

Start with the case study, follow the evidence link, inspect the primary repository or report, and then review the stated limitation.

Open evidence hubView CVContact