AI & Data Science
The fastest way to understand a model is to compute it by hand once. In these courses you build machine-learning methods step by step in a spreadsheet: every weight, distance, split and error is a cell you can inspect. The checker confirms each step, so by the end you know what the library function in your code is actually doing.
The 25 courses cover classic machine learning, neural networks, decision trees and ensembles, clustering, text classification, recommendation, embeddings and vector search, reinforcement learning, anomaly detection and feature engineering. Practical tracks deal with evaluating AI systems, fairness metrics, model monitoring, labelling quality, GPU capacity, token costs and the business case for automation.
AI & Data ScienceScaling, regression, gradient descent, logistic regression, precision/recall, kNN, k-means and decision trees, computed cell by cell on small data, so you see what Python libraries do inside.
AI & Data ScienceNeurons, activation functions, a full forward pass, loss, backpropagation, learning rates, softmax and early stopping, computed cell by cell so you see what PyTorch and TensorFlow do inside.
AI & Data ScienceRatings matrices, cosine similarity, user- and item-based collaborative filtering, MAE against a popularity baseline, bag of words, TF-IDF and lexicon sentiment, built cell by cell for one online bookshop.
AI & Data SciencePrice LLM requests from tokens, forecast the monthly AI bill, score models on cost, quality and speed, value automation savings, and build an ROI, payback and risk-adjusted business case, on live auto-graded spreadsheets.
AI & Data ScienceMeasure search, RAG retrieval and LLM quality hands-on: precision and recall@k, MRR, NDCG, MAP, ROC curves and AUC, calibration error, LLM-judge agreement with Cohen's kappa and A/B win rates with confidence intervals, every cell checked live.
AI & Data ScienceModel what AI features really cost: tokens per request and per chat, cost per 1,000 requests, usage tiers, prompt caching and batch discounts, model routing, RAG context, cost per resolved ticket vs people, and gross margin by plan, on live auto-graded spreadsheets.
AI & Data ScienceBuild the core RL algorithms cell by cell with given random numbers: bandit estimates, ε-greedy and UCB, discounted returns, the Bellman equations, value iteration, Q-learning updates and policy extraction, every cell checked live.
AI & Data ScienceEvaluate an object detector cell by cell: box formats and YOLO coordinates, IoU with MAX and MIN, matching predictions to ground truth, precision and recall at thresholds, the PR curve, 11-point and all-point AP, mAP and mAP@[.5:.95], class confusion and Dice vs IoU for masks, every cell checked live.
AI & Data ScienceAudit a (fictional) loan-approval model hands-on: confusion matrices by group, demographic parity, the disparate impact ratio and four-fifths rule, equal opportunity and equalised odds, calibration by group, threshold trade-offs, intersectional groups and a flagged fairness report card, every cell checked live.
AI & Data ScienceFlag fraud and outages the way monitoring and risk teams do: z-scores, robust z with median and MAD, IQR fences, rolling ±kσ bands, seasonal baselines, multivariate distance, precision/recall against labelled incidents and the cost of alert fatigue, in 8 graded labs.
AI & Data ScienceForecast daily demand the way ML teams do: lag, rolling and calendar features, a leak-free split by date, naive and seasonal-naive baselines, MAE/RMSE/MAPE, a LINEST regression on lags, walk-forward validation and drop-one feature importance, in 8 graded labs.
AI & Data ScienceBuild a Naive Bayes text classifier cell by cell: priors, word counts with SUMIFS, Laplace smoothing, log-probability scores with SUMPRODUCT, a confusion matrix, precision/recall/F1 at a log-odds threshold, a lexicon-vs-Naive-Bayes sentiment test and the most informative words, in 8 graded labs.
AI & Data ScienceBuild a credit-default tree cell by cell: entropy and Gini, information gain with COUNTIFS, the best numeric cut-point, a two-level tree as nested IF, overfitting and pruning, a random-forest vote with out-of-bag rows, gradient boosting with residuals, and feature importance, in 8 graded labs.
AI & Data ScienceVector length and normalisation, dot product vs cosine with SUMPRODUCT, top-k retrieval with RANK.EQ, LARGE and INDEX/MATCH, Euclidean vs cosine ordering, chunking cost, ANN recall@k, hybrid keyword + vector scoring and RAG retrieval evaluation with hit rate, MRR and a chart: 8 labs on tiny embeddings you can read.
AI & Data ScienceIntent mix, containment and escalation rates, deflection vs true resolution with the 7-day re-contact test, thumbs-up CSAT, cost per resolution bot vs human, handoff quality and an intent-level ROI roadmap: 8 labs on one fictional SaaS support bot.
AI & Data ScienceDistance matrices, single vs complete linkage, dendrogram cuts, silhouette scores, elbow vs silhouette, a PAM swap and DBSCAN core/border/noise, every step computed cell by cell on 12 stores.
AI & Data ScienceScore a speech-to-text engine hands-on: WER from alignment counts, a word-by-word comparison table, the Levenshtein edit-distance table, character error rate with MID, WER by accent and noise, real-time factor and latency percentiles, custom-vocabulary impact and cost per audio hour, in 8 graded labs.
AI & Data ScienceRun a genetic algorithm cell by cell with given random numbers: encode solutions as bits, score them with penalties, spin the roulette wheel, hold tournaments, cross and mutate chromosomes, step a full generation and check convergence against the brute-force optimum.
AI & Data ScienceModel memory for FP16/INT8/INT4, KV-cache per token, throughput vs batch size, time to first token, GPUs for the peak with Little's law, cost per 1M tokens, self-hosting vs API break-even and a six-month GPU plan: 8 graded labs on one fictional AI start-up.
AI & Data ScienceMonitor a (fictional) credit-scoring model hands-on: bin baseline and current data, calculate the PSI, the Kolmogorov–Smirnov statistic, chi-square and Jensen–Shannon drift, track prediction drift and performance decay once labels arrive, build a traffic-light alert dashboard and decide when retraining pays, every cell checked live.
AI & Data ScienceMeasure and budget the labels behind an AI model in a live grid: class balance and weights, percent agreement vs Cohen's kappa, Fleiss' kappa, majority vote and confidence, gold-item accuracy, label-noise effects, labelling cost per item and active learning savings, every cell checked as you go.
AI & Data ScienceTreat images as numeric grids: greyscale from RGB, brightness, contrast and FREQUENCY histograms, thresholds and masks, 3×3 convolution with SUMPRODUCT, sharpening and Sobel edges, padding and stride, max and average pooling, and a tiny CNN from conv to probability, 8 graded labs with colour-scale heatmaps.
AI & Data SciencePrepare a loan-default dataset the way ML teams do: missing-value flags and group imputation, one-hot, count and ordinal encoding, min-max, z-score and robust scaling, quantile bins, log transforms and winsorising, ratio and interaction features, smoothed target encoding without leakage, and a correlation + AUC feature screen, in 8 graded labs.
AI & Data ScienceNormalise messy company names, block candidate pairs, score them with bigram and token similarity, combine the evidence into match / review / non-match decisions, measure precision and recall, and merge duplicates into clusters and golden records, in a live spreadsheet that checks every cell.
Budget an AI training run before you start it: dataset tokens, compute, GPU-hours, fine-tuning costs, experiment grids and scaling curves, with every cell checked as you go.
Confusion matrix, precision and recall with an Excel example: TP, FP, FN and TN from COUNTIFS, then accuracy, F1, specificity and the threshold trade-off.
Calculate LLM API costs in Excel: input and output tokens × price per million, cost per request, per 1,000 requests and per month, plus caching savings.
NDCG and MRR explained with examples: discounted gain, the ideal ranking, NDCG@k, reciprocal rank and MRR, calculated step by step in an Excel table.
The intersection over union (IoU) formula for bounding boxes: overlap with MAX and MIN, union, the 0.5 threshold, Dice vs IoU. Worked example.
How to calculate the disparate impact ratio: selection rates, impact ratios, the four-fifths (80%) rule and a significance check. Worked example.
How to calculate entropy and information gain for a decision tree split, step by step, with Gini, gain ratio and an Excel example on 20 loans.
No. You work in a spreadsheet, which makes every calculation visible. It complements coding courses well: once you understand the mechanics, the code is easier to read.
School-level algebra is enough to start. Each course introduces the formulas it needs and lets you check your understanding cell by cell.
Each course is a single purchase of $9.99. That covers every lesson and lab, the final exam and the certificate, with no subscription and nothing that renews.
Start with the recommended course and build your first result in the lab today.
