Beginner to intermediate · Course
The 15-Day AI/ML Interview Bootcamp
Machine learning from first principles to the interview room — one lesson and one assignment a day.
Your team "added AI" this year. Now every interview you sit has a round you didn't sign up for: "How does gradient descent work?", "Precision or recall here — and why?", "Explain attention like I'm a backend engineer." You've read the blog posts. You've watched the videos. And you still can't build one of these things from a blank file, which is exactly what the room wants to see.
That's this course. Fifteen days, one idea a day, every idea built by hand. No libraries, no notebooks you can't explain — each algorithm is written from scratch in tested code, drawn out before it's coded, and tied straight back to the question an interviewer will ask you about it. Every night ends with an assignment that has exact numbers to hit, and every next morning opens with the answer key.
What's inside
- Week 1 — The learning loop and classical ML: what learning actually is, data and the split you must never get wrong, linear regression and gradient descent, logistic regression and the metrics that matter, overfitting and regularisation.
- Week 2 — Trees to transformers: decision trees and ensembles, Naive Bayes and text, k-means and PCA, neural networks and backpropagation, training deep nets well, CNNs and RNNs, attention and the transformer block.
- Days 13–15 — LLMs, ML systems, and the mock round: tokens, prompting, embeddings and RAG; ML system design with a real serving endpoint; and a full timed mock interview with a grading rubric.
By Day 15 you'll own a small machine-learning library you wrote yourself — and, more usefully, the ability to explain every line of it under pressure.
Day 1 is free. Build your first model tonight, see how the method feels, and if it clicks, the rest of the bootcamp is one small payment away — yours for life, alongside every other course and article on the site.
What you’ll be able to do
- Explain how a model actually learns — loss, gradients, and the update loop — on a whiteboard, from memory.
- Implement linear and logistic regression, k-means, a decision tree, and a neural network from scratch, and defend every line.
- Diagnose overfitting, pick the right metric, and design a validation scheme the interviewer can't poke a hole in.
- Walk through attention, transformers, and LLM/RAG systems at the depth a senior engineer is expected to.
- Run an ML system-design round end to end — data, features, training, serving, monitoring — and finish a timed mock.
Why the course?
The deep-dive articles on this site are free — the course saves you the one thing you’re short on before an interview: time.
One week, one path
No thousand-tab rabbit hole. A day at a time, in the order that actually builds on itself.
Tested, not hand-wavy
Every snippet compiles and runs, and the traps that fail the round are called out so you don’t fall in.
Yours for life
Come back before every interview you’ll ever sit. One payment, no subscription.
The 15-day plan
One topic a day. Day 1 is free — start there.
- Day 1 — What machine learning actually is, and your first model Free
AI vs machine learning vs deep learning vs generative AI, the learning loop every model shares (data → model → loss → update), supervised vs unsupervised vs reinforcement learning, and a k-nearest-neighbours classifier built from scratch — with the interview questions that open every ML round.
- Day 2 — Data, features, and the split you must never get wrongPremium
Train, validation and test splits, data leakage, stratification, feature scaling, one-hot encoding, missing values, and the tiny Matrix and Dataset utilities the rest of this machine learning course is built on — with the data questions interviewers use to separate practitioners from readers.
- Day 3 — Linear regression and gradient descent, from a straight line to the loss landscapePremium
Fit a line with least squares, define mean squared error, walk down the loss surface with gradient descent, pick a learning rate that converges, and compare with the closed-form normal equation — machine learning built from scratch, with the gradient descent interview questions.
- Day 4 — Logistic regression, classification metrics, and the threshold nobody tunesPremium
Logistic regression from scratch: the sigmoid, cross-entropy loss, then confusion matrix, accuracy, precision, recall, F1, ROC and AUC — and why 99% accuracy can be a failing model. With the classification-metrics questions every machine learning interview asks.
- Day 5 — Overfitting, the bias–variance trade-off, and regularisationPremium
Why a model that aces training data fails in production, the bias-variance trade-off drawn rather than stated, L1 and L2 regularisation from scratch, k-fold cross-validation, learning curves, and how to tell 'more data' from 'simpler model' — the single most asked machine learning interview topic.
- Day 6 — Decision trees, random forests, and gradient boostingPremium
Grow a decision tree from scratch with Gini impurity and information gain, see why a single tree overfits, then fix it two ways — bagging into a random forest and boosting into gradient-boosted trees — the ensemble questions every machine learning interview asks.
- Day 7 — Naive Bayes and your first text classifierPremium
Bayes' theorem in plain words, the naive independence assumption and why it works anyway, tokenisation, bag of words, TF-IDF, Laplace smoothing, and a spam classifier built from scratch — the probability and NLP basics every machine learning interview tests before you talk about transformers.
- Day 8 — k-means clustering and PCA, learning without labelsPremium
Unsupervised machine learning from scratch — Lloyd's k-means algorithm, k-means++ initialisation, choosing k with the elbow and silhouette, then principal component analysis by power iteration to compress and visualise data — with the clustering questions interviewers probe.
- Day 9 — Neural networks from scratch: perceptron to backpropagationPremium
Build a neural network from a single perceptron up — layers, activations, the forward pass, the chain rule, backpropagation by hand and in code, and stochastic gradient descent — solving XOR from scratch, with the backpropagation questions every machine learning and deep learning interview asks.
- Day 10 — Training deep networks: initialisation, Adam, dropout, and early stoppingPremium
Why deep networks stall — vanishing gradients — and the toolkit that fixes it: Xavier and He initialisation, momentum and Adam, batch normalisation, dropout, learning-rate schedules and early stopping, built from scratch — with the machine learning interview questions on training dynamics.
- Day 11 — Convolutional networks and recurrent networks, the ideas behind themPremium
Convolution as a sliding filter, pooling, weight sharing and why CNNs beat dense layers on images; recurrence, hidden state and LSTM gates for sequences; both implemented from scratch on tiny inputs, with the CNN and RNN questions interviewers use to check you understand the structure.
- Day 12 — Attention and the transformer, from scratchPremium
Embeddings, queries, keys and values, scaled dot-product attention, multi-head attention, positional encoding, the encoder block with residuals and layer norm, and why the transformer replaced the RNN — built from scratch, with the attention questions every machine learning interview asks.
- Day 13 — LLMs in practice: tokenisation, prompting, embeddings, vector search, and RAGPremium
How large language models are trained and used — tokenisation, context windows, sampling, prompting, embeddings and cosine similarity, vector search, retrieval-augmented generation, hallucination and evals — with a mini RAG pipeline built from scratch and the LLM questions every AI interview asks.
- Day 14 — ML system design: from notebook to production, with a Spring Boot serving endpointPremium
The machine learning system design interview end to end — requirements, feature pipelines and feature stores, offline vs online evaluation, A/B tests, serving latency, monitoring, drift and retraining — on a recommender, then a Spring Boot endpoint that serves the Day 4 fraud model.
- Day 15 — The AI/ML mock interview: coding, concepts, and the case studyPremium
A full AI/ML mock interview — the 40 concept questions with strong one-line answers, the three from-scratch coding tasks interviewers actually set (k-means, logistic regression, attention), the ML case-study framework, a grading rubric, and the one-page revision sheet for the night before.