01Data Analytics
From raw data to a defensible story
A repeatable workflow for turning an unfamiliar dataset into clear, evidence-based findings without losing sight of data quality.
LevelBeginnerTime45–60 minFormatPython · pandas
01Frame the question
Write one decision question, define the unit of analysis and state what a useful answer must contain before opening the data.
Audit before analysing
Check shape, data types, missingness, duplicates, ranges and category consistency. Record every correction rather than silently cleaning.
Build the evidence chain
Move from distributions to comparisons, then visualise only the relationship that answers the original question. Separate observations from interpretation.
OutcomeA clean analysis table, three validated findings and one decision-ready visual.
Try itChoose a public CSV. Produce a one-page note containing the question, five quality checks, three findings, one chart and two limitations.
02Cybersecurity
Map a threat to testable controls
A safe, structured method for moving from a realistic threat scenario to evidence-based security controls and a verification plan.
LevelIntermediateTime40–50 minFormatThreat model · CVE research
01Define scope and rules
Identify the system boundary, valuable assets, likely threat actors and explicit rules of engagement. Never test outside authorised scope.
Connect weakness to impact
Map each plausible weakness to a relevant attack path, affected asset and business consequence. Use authoritative vulnerability sources.
Specify control evidence
For every preventive or detective control, define the evidence that would prove it works and the condition that would trigger remediation.
OutcomeA compact threat-to-control matrix with evidence and re-test criteria.
Try itCreate a matrix for a fictional small web service: three threats, three controls, an evidence source for each control and a measurable re-test condition.
03Research Methods
Design a reproducible experiment
Translate a broad research idea into a transparent study that another researcher can inspect, reproduce and challenge.
LevelAll levelsTime50–70 minFormatResearch canvas · Experiment log
01Narrow the claim
Convert the topic into a falsifiable question. Name the population, intervention or method, comparison and measurable outcome.
Lock the evaluation
Choose data splits, baselines, metrics and stopping rules before examining final results. Explain why each metric fits the decision.
Plan for scrutiny
Record seeds, versions, preprocessing, exclusions and limitations. Add one ablation and one robustness test that could disconfirm the claim.
OutcomeA one-page experimental protocol with variables, baselines, metrics and validity checks.
Try itDraft a protocol for comparing two anomaly detectors on an imbalanced dataset. Include leakage controls, at least three metrics and one robustness test.