AMLSS · 2026/
Amazon
Next chapter/Applied Scientist Intern/2026 cohort

You finished the school.Now claim the offer.

Congratulations on completing Amazon ML Summer School. The learning phase is complete, and the Applied Scientist Intern opportunity is the next step. The process may take up to 2 months from here, so please be patient. Timelines, rounds, and requirements can vary across years and hiring cycles. Use this guide as a reference, not a fixed process.

Program completeDSA + ML interviewBuild depth, not just recall
Start the roadmapOfficial guide
Applied Scientist interview preparation
From MLSS to InternAMLSS · 2026

The path to an offer

Every round is eliminatory. Prepare each one as a connected hiring, not an isolated test.

STAGE · 01

Resume & interest

Make every project, metric, and technical choice on the page defensible.

STAGE · 02

Online assessment

Expect two medium DSA problems. Clear, efficient reasoning matters as much as passing tests.

STAGE · 03

Interview Round 1: DSA

Prepare LeetCode medium problems and behavioural questions. Start with a baseline, then explain the optimisation, time complexity, space complexity, and edge cases.

STAGE · 04

Interview Round 2: ML

Expect depth and breadth across ML fundamentals, your projects, model choices, datasets, metrics, and real-world trade-offs.

Resources worth returning to

Use the official references for framing, then turn the material into your own explanations.

Official interview preparation

Amazon's guidance on preparation, process, and interview expectations.

Open resource

Machine Learning playlist

Build an organised revision sequence for core machine-learning concepts.

Open resource

ML interview questions

Use this question bank to practise concise explanations and follow-ups.

Open resource

Leadership Principles

Build natural STAR stories around the principles that shape Amazon behavioural interviews.

Open resource

Transformers documentation

A clear reference for causal and masked language modeling fundamentals.

Open resource

Amazon DSA practice

Use tagged questions to practise patterns instead of collecting random solutions.

Open resource
Reference notes

Review the program slides.

Revisit the Amazon notes from the program whenever you need a focused revision.

Interview Round 1 · DSA

Prepare for a real coding interview.

  • Start from a brute-force baseline, then verbalise the exact bottleneck before optimising.
  • Be fluent with arrays, strings, linked lists, trees, graphs, recursion, DP, stacks, queues, hashing, and binary search.
  • For every final solution, state time complexity, auxiliary space, mutation assumptions, and edge cases.
  • Practise linked resources in the candidate notes below: Number of Islands, 3Sum, Maximum Subarray, Minimum Window Substring, anagrams, and tree distance.
Interview Round 2 · ML depth & breadth

Machine Learning Concepts & your resume.

  • Build precise explanations for supervised and unsupervised learning, bagging, boosting, bias–variance, losses, trees, clustering, and evaluation.
  • Expect deep learning follow-ups: backpropagation, optimisation, dropout, CNNs, RNNs/LSTMs, transformers, attention, and normalization.
  • Your major and minor projects are interview material: expect detailed cross-questions on accuracy, alternative models, data collection, training/testing, features, database choices, failure modes, and why one approach was chosen over another.
  • If you mention research work or a paper, revisit every method, baseline, result, limitation, and possible extension. Interviewers can ask about any part of it.
  • Refresh probability, statistics, linear algebra, entropy, Gini, Bayes' theorem, and metric choices under class imbalance.

Build pattern fluency, not a question collection

For every solution: brute force → bottleneck → optimal idea → complexity → edge cases.

01Construct Binary Tree02Binary Tree Maximum Path Sum03Serialize & Deserialize Binary Tree04Number of Islands05Course Schedule II06Minimum Window Substring07Top K Frequent Words08Reverse Nodes in k-Group09Merge k Sorted Lists10Trapping Rain Water

Also cover 1D/2D DP, memoization, tabulation, state definition, transitions, and space optimisation through House Robber, Coin Change, LCS, and LIS.

Answer like a scientist

A dependable structure that makes technical follow-ups easier to handle.

  1. 01

    Define it

    State what the concept is, precisely and without jargon.

  2. 02

    Explain the intuition

    Describe why it works before moving into the mechanics.

  3. 03

    Go inside the model

    Cover the objective, calculation, architecture, or algorithmic steps.

  4. 04

    Ground it

    Give a practical project or product example where you would use it.

  5. 05

    Name the trade-off

    Explain when you would not use it and what you would choose instead.

ML essentials

Know the follow-up.

  • Classical ML: split criteria, entropy, Gini, information gain, bagging vs boosting, Random Forest, XGBoost, KNN, K-Means, PCA, regularisation, bias and variance.
  • Metrics: accuracy, precision, recall, F1/F-beta, ROC-AUC, PR-AUC, thresholds, confusion matrices, and the right choice for imbalanced data.
  • Probability: Bayes' theorem, conditional probability, independence, expectation, variance, covariance, Bernoulli/Binomial distributions, MLE, and basic hypothesis testing.
Deep learning & transformers

Understand the mechanics.

  • Explain forward propagation, backpropagation, losses, ReLU/sigmoid/softmax, SGD vs Adam, learning rate, batch size, dropout, BatchNorm, LayerNorm, and initialization.
  • Trace one token: tokenization → embeddings → position → Q/K/V → attention → multi-head attention → add & norm → feed-forward → residual layers → prediction.
  • Be ready for BERT vs GPT, masked vs causal language modeling, encoder vs decoder, autoregressive generation, O(N²) attention, and why Sparse or Flash Attention helps.
From model to production

Keep systems practical.

  • Understand the lifecycle: data → cleaning → features → training → validation → deployment → inference → monitoring → retraining.
  • Practise concise designs for a movie recommender, spam detector, or product recommender—cover batch vs real-time inference, latency, throughput, feedback loops, and cold start.
  • Prioritise ML fundamentals and project depth before spending weeks on infrastructure tools. For this internship, systems are a secondary layer, not the starting point.

A focused 30-day preparation rhythm

Use it as a baseline and rebalance it around your weakest areas, projects, and upcoming interview date.

STAGE · W1

Foundations

Arrays, strings, hashing, two pointers, sliding window, binary search, linked lists; supervised learning, regression, classification, metrics, bias–variance.

STAGE · W2

Structure

Trees, graphs, BFS/DFS, stacks, queues, heaps, basic DP; trees, forests, boosting, K-Means, PCA, and regularisation.

STAGE · W3

Depth

Neural networks, backpropagation, optimisation, dropout, normalisation, CNNs, RNNs/LSTMs, transformers—and a line-by-line resume review.

STAGE · W4

Interview mode

Timed DSA aloud, ML answers aloud, probability drills, project mock interviews, and STAR stories for leadership principles.

What candidates actually saw

Interview experiences from Amazon ML Summer School (2023-2025)

A practical final-week checklist

This process can take one to two months. Keep learning, stay patient, and prepare steadily.

Build clear stories for your projects: explain the problem, your approach, why you made each technical decision, the result, and what you would improve next.
Practice DSA under time pressure. Explain your approach aloud, start with a baseline solution, optimize it, and always state the time and space complexity of the solution you implement.
Revise the core ML fundamentals: loss functions, evaluation metrics, Bayes theorem, entropy, gradient descent, regularization, bias-variance, trees, ensembles, and neural networks.
Prepare examples around Amazon Leadership Principlesusing STAR: Situation, Task, Action, and Result. Focus on what you personally did and the measurable outcome.
Rehearse behavioural questions about learning quickly, handling failure, receiving negative feedback, resolving disagreements, working with incomplete information, and taking ownership of difficult problems.
Know your resume inside out. Be ready to explain every project, technology, dataset, model, metric, and result. If you mention something, expect the interviewer to ask "Why?" and follow up several times.
Prepare for ML depth. For the models and techniques you have used, understand the intuition, mathematics, implementation, and trade-offs instead of memorizing definitions.
Communicate your reasoning clearly during live coding. If you do not immediately reach the final solution, explain your assumptions, approach, dry run, edge cases, and complexityas you work toward the solution.