/knowledge

What I learned, and keep learning.

Thorough, first-principles explainers of the data science I studied at the University of Melbourne — and the topics I taught. Writing each one from scratch is how I keep the fundamentals sharp. Foundation first, then advanced; built one topic at a time.

68 of 68 live

Foundation

16/16

The maths, statistics, data and systems base — Bachelor of Science.

Linear AlgebraVectors, matrices, eigenvectors, the SVD

ProbabilityRandom variables, distributions, Bayes' rule

StatisticsEstimation, inference, hypothesis tests

Calculus & OptimisationGradients, gradient descent, backprop

Linear Statistical ModelsOLS, inference, diagnostics

Database SystemsRelational model, SQL, indexing

Artificial IntelligenceSearch, logic, planning

Web Information TechnologyHTTP, HTML/CSS/JS, REST APIs, the request cycle

Operations ResearchLinear programming, feasible regions, the simplex method

Elements of Data ProcessingThe pipeline, tidy data, cleaning, reshaping, joins

Applied Data ScienceThe project lifecycle, CRISP-DM, problem-first

Data Visualisation & PerceptionEncoding accuracy, pre-attentive cues, honest charts

Feature Engineering & Data PreparationCleaning, scaling, encoding, and avoiding leakage

Sampling & Survey MethodologyFrames, probability sampling, bias, weighting

SQL & Querying DataExecution order, joins, window functions, NULLs

Model Evaluation & ValidationCross-validation, ROC/AUC, the metric that matters

Advanced

34/34

Built on the foundation — Master of Data Science.

Natural Language ProcessingTokens, TF-IDF, embeddings, transformers

Statistical Machine LearningBias-variance, regularisation, generalisation

Bayesian StatisticsPriors, posteriors, MCMC

PCA & Dimensionality ReductionCovariance, eigenvectors, variance explained

Clusteringk-means, hierarchical, choosing k

Cluster & Cloud ComputingMPI, Spark, HPC at scale

Statistical ModellingGLMs, logistic & Poisson, link functions, AIC

Computational StatisticsMonte Carlo, the bootstrap, permutation tests, EM

Advanced Database SystemsQuery optimisation, MVCC, distribution, columnar

Science Communication at WorkBriefing decisions: lead with the recommendation

Time Series AnalysisTrend, seasonality, ARIMA, forecasting honestly

Causal Inference & A/B TestingCounterfactuals, RCTs, confounders, DiD

Deep Learning & Neural NetworksNeurons, backprop, gradient descent, CNNs to transformers

Reinforcement LearningReward, MDPs, Bellman, Q-learning, reward hacking

Ensemble Methods & Gradient BoostingBagging, random forests, boosting, XGBoost

Recommender SystemsCollaborative filtering, matrix factorisation, cold start

Survival AnalysisCensoring, Kaplan-Meier, Cox hazard ratios

Information Retrieval & SearchInverted index, BM25, dense & hybrid retrieval

Large Language ModelsNext-token prediction, RLHF, hallucination, RAG

Topic ModellingLDA, themes without labels, choosing topics

AI Agents & Tool UseThe reason-act loop, tools, planning, verification

Spatial StatisticsAutocorrelation, Moran's I, hotspots, kriging

Conformal Prediction & UncertaintyGuaranteed coverage intervals around any model

Causal DiscoveryLearning the DAG, PC & GES, Markov equivalence

State-Space Models & the Kalman FilterHidden state, predict-update, the Kalman gain

Active & Semi-Supervised LearningLearning with few labels, uncertainty sampling

Extreme Value TheoryTail risk, GEV/GPD, return levels, the 1-in-N event

Hierarchical & Mixed-Effects ModelsGrouped data, partial pooling, shrinkage

Optimisation MethodsConvex vs not, LP/simplex, integer, metaheuristics

Gaussian ProcessesDistribution over functions, kernels, uncertainty

Robust StatisticsBreakdown point, median/MAD, M-estimators, Huber

Quantile RegressionConditional quantiles, pinball loss, the whole spread

Graph Neural NetworksMessage passing, node embeddings, GCN/SAGE/GAT

Probabilistic Graphical ModelsBayes nets, factorisation, inference under uncertainty

In Practice

16/16

What I do now — applied to current government-analyst work.

Business Intelligence & DashboardsPower BI, star schema, one question per page

Geospatial Analysis & GISSpatial joins, choropleths, rates, the MAUP

Intelligence Analysis & OSINTThe intelligence cycle, ACH, source grading, bias

Data Governance, Privacy & EthicsFAIR, privacy, access, fairness, lineage

Anomaly DetectionOutliers, isolation forest, alert-fatigue trade-off

Network & Graph AnalysisCentrality, brokers, communities, link analysis

Reproducibility & Analytical PipelinesVersion control, environments, RAP, lineage

MLOps & Model MonitoringDeployment, drift, the retrain loop, skew

Explainable AI & InterpretabilitySHAP, LIME, glass-box models, defensible calls

Fairness & Bias in MLProxies, fairness metrics, the impossibility result

Differential PrivacyRe-identification, epsilon, calibrated noise, utility

Knowledge GraphsTriples, ontologies, entity resolution, GraphRAG

Streaming & Real-Time AnalyticsWindowing, event time, watermarks, Kafka/Flink

Data Architecture & the Modern StackWarehouse/lake/lakehouse, ELT, dbt, medallion

Statistical Process ControlCommon vs special cause, control charts, ±3σ

Federated LearningTrain without sharing data, FedAvg, non-IID

Taught

2/2

Topics I taught or mentored — written from the other side of the desk.

Data Science mentoring themesUniMelb peer mentoring, 2024

EdTech & the Science of LearningCognitive load, spacing, retrieval, desirable difficulties