Practice
Implement the primitives that ML engineering interviews and real systems are built on. Every problem runs in your browser. No setup, no submission queue — just write Python and press Run.
Featured
The 100 most interview-relevant ML problems — attention, backprop, optimizers, metrics & more — hand-picked from the catalog and ranked.
Build intuition from scratch — problems beyond the interview set, chosen to pair with the step-by-step, animated Code Explainer.
Prep by company
GoogleSearch, Ads, DeepMind, YouTube. Strong on classical ML, ranking, and large-scale serving.
MetaMeta AI / Reality Labs / Ads ranking. Heavy on PyTorch internals, recommendation systems, multi-task learning.
OpenAILLM training and post-training. Expect transformer internals, attention variants, sampling, and tokenization.
AnthropicFrontier LLMs and safety. Transformer mechanics, RLHF/RLAIF, evals, and inference-time techniques.
- #1Accuracy scoreEasyEvaluation Metrics10 tests
- #2AdaBoost fit (decision stumps)MediumEnsemble LearningBoosting10 tests
- #3Adadelta optimizerMediumCalculus10 tests
- #4Adagrad optimizerEasyCalculus10 tests
- #5Adam optimiser stepMediumCalculus10 tests · Meta · OpenAI · Anthropic
- #6Adamax optimizerMediumCalculus10 tests
- #7AdamW step (decoupled weight decay)MediumCalculusRegularization10 tests
- #8Affine coupling layer (normalizing flow)MediumNeural Networks10 tests
- #9Assemble the Critic's audit inputEasyAgentic AI10 tests
- #10Build the chat messages listEasyLLMsAgentic AI10 tests
- #11Cache-before-search, write-after-fetchMediumAgentic AI10 tests
- #12Which ReAct step are we on?EasyAgentic AI10 tests
- #13PASS or FAIL?EasyEvaluation MetricsAgentic AI10 tests
- #14Count the dedup savingsEasyAgentic AI10 tests
- #15Find the bottleneck executorMediumML System DesignAgentic AI10 tests
- #16Format retrieved docs as contextEasyLLMsML System DesignAgentic AI10 tests
- #17Map your components to the frameworksEasyAgentic AI10 tests
- #18Rebuild findings from gather resultsEasyAgentic AI10 tests
- #19One model, three roles — route by promptEasyLLMsAgentic AI10 tests
- #20Accumulate findings across attemptsEasyAgentic AI10 tests
- #21MockLLM keyword routingMediumLLMsAgentic AI10 tests
- #22MockSearchEngine keyword retrievalEasyLLMsML System DesignAgentic AI10 tests
- #23The Orchestrator feedback loopMediumAgentic AI10 tests
- #24Parallel execution profileMediumAgentic AI10 tests
- #25Parse the planner's sub-question listEasyAgentic AI10 tests
- #26Parse the SEARCH / ANSWER protocolEasyAgentic AI10 tests
- #27Parse the Critic's structured verdictMediumAgentic AI10 tests
- #28Agents are nouns; these are the verbsEasyAgentic AI10 tests
- #29Plan → execute → synthesizeMediumAgentic AI10 tests
- #30The RAG run: search → ground → answerMediumLLMsML System DesignAgentic AI10 tests
- #31The ReAct loop with a max_steps guardMediumLLMsAgentic AI10 tests
- #32The whole system, expressedMediumAgentic AI10 tests
- #33The SharedMemory storeEasyAgentic AI10 tests
- #34When does the retry loop stop?EasyAgentic AI10 tests
- #35Aggregate trace eventsEasyML System DesignAgentic AI10 tests
- #36Run totals from the traceEasyML System DesignAgentic AI10 tests
- #37Validate a chat messageEasyLLMsEvaluation MetricsAgentic AI10 tests
- #38Sequential vs parallel wall-clockEasyAgentic AI10 tests
- #39Add & Norm (post-norm residual)MediumTransformersNormalization10 tests
- #40Fuse token + positional embeddingsEasyTransformersEmbeddings10 tests
- #41Attention step 3 — weighted sum of valuesEasyTransformers10 tests
- #42Teacher forcing — next-token targetsMediumTransformersNLPNeural NetworksLLMs10 tests
- #43Concatenate the headsEasyTransformers10 tests
- #44Cross-attention — decoder Q, encoder K/VMediumTransformers10 tests
- #45Cross-attention score shape (target × source)EasyTransformers10 tests
- #46One decoder layer (post-norm, 3 sublayers)MediumTransformers10 tests
- #47How many sublayers per layer?EasyTransformers10 tests
- #48Head dimension d_k = d_model / hEasyTransformers10 tests
- #49The input layer — token lookup + positionalMediumTransformersEmbeddings10 tests
- #50Per-layer complexity — attention vs recurrenceMediumTransformersRNN10 tests
Meta
OpenAI
AnthropicShowing 1–50 of 468