Microsoft

Solve a problem (pass its tests) and it's checked off here automatically — your progress is saved, so it stays checked when you come back.

• The RAG run: search → ground → answerLargeLanguageModelMLSystemDesignAgenticAIMedium
• Full self-attention from XTransformersMedium
• Which transformer family?TransformersEasy
• Bayesian A/B Test (Beta-Binomial)ProbabilityStatisticsMedium
• Bootstrap Confidence Interval (Percentile)StatisticsMedium
• Brier Score + Reliability DecompositionEvaluationMetricsMedium
• Conv2D forward (naive, stride 1, no pad)ConvolutionNeuralNetworkMedium
• Conv2D forward (padding + stride)ConvolutionNeuralNetworkMedium
• Scaled dot-product self-attentionTransformersLargeLanguageModelMedium
• Causal mask: build + applyTransformersLargeLanguageModelEasy
• LayerNorm forwardTransformersLargeLanguageModelNormalizationEasy
• Chi-Square Test of IndependenceStatisticsMedium
• Clipped Inverse Propensity WeightingCalculusMLSystemDesignMedium
• CUPED Variance ReductionStatisticsMedium
• Delta Method for Ratio Metric VarianceStatisticsHard
• Difference-in-DifferencesStatisticsMedium
• Overfitting — read the learning curvesNeuralNetworksOverfittingRegularizationEasy
• Expected Calibration Error (ECE)EvaluationMetricsMedium
• Expected Reciprocal Rank (ERR)MLSystemDesignEvaluationMetricsMedium
• F1 score (binary classification)EvaluationMetricsEasy
• Generate random subsets (bootstrap)StatisticsEnsembleLearningBaggingEasy
• Pre-LN vs post-LN blockTransformersConvolutionNeuralNetworkLargeLanguageModelMedium
• Gradient Boosting: One Round (Stump)DecisionTreesEnsembleLearningBoostingMedium
• Hash Bucket A/B AssignmentStatisticsEasy
• Histogram Split Finding (GBM)DecisionTreesEnsembleLearningBoostingHard
• Holt-Winters Exponential SmoothingStatisticsMedium
• Team-Draft Interleaving (Online Ranking Eval)StatisticsMLSystemDesignMedium
• Isotonic Regression (PAVA)EvaluationMetricsHard
• Jensen-Shannon Divergence (Drift)ProbabilityMLSystemDesignMedium
• K-means: one iterationStatisticsMedium
• Knowledge-Distillation LossMLSystemDesignLossFunctionsFineTuningMedium
• Kolmogorov-Smirnov StatisticStatisticsMLSystemDesignMedium
• LambdaMART Lambda GradientsMLSystemDesignEvaluationMetricsHard
• ListNet Listwise Ranking LossMLSystemDesignLossFunctionsMedium
• Logistic regression — gradient descentLinearRegressionLogisticRegressionCalculusMedium
• Logistic regression — binary cross-entropy lossLinearRegressionLogisticRegressionLossFunctionsEasy
• Logistic regression — probability from a scoreLinearRegressionLogisticRegressionActivationFunctionsEasy
• Longest Common SubsequenceNaturalLanguageProcessingMedium
• Mann-Whitney U TestStatisticsMLSystemDesignMedium
• Mean Average Precision (MAP)MLSystemDesignEvaluationMetricsMedium
• Multiple-Testing Correction (Bonferroni / BH-FDR)StatisticsMedium
• pass@k Unbiased EstimatorStatisticsLargeLanguageModelEvaluationMetricsMedium
• Platt ScalingLogisticRegressionEvaluationMetricsMedium
• Population Stability Index (PSI)StatisticsMLSystemDesignMedium
• PPO clipped objectiveLargeLanguageModelFineTuningAgenticAIMedium
• Recursive Chunking with OverlapNaturalLanguageProcessingLargeLanguageModelMLSystemDesignMedium
• ROC-AUC from scratchEvaluationMetricsMedium
• Sample Ratio Mismatch (SRM) CheckStatisticsMLSystemDesignMedium
• Sample Size for Statistical PowerStatisticsMedium
• Self-Consistency Majority VoteLargeLanguageModelEvaluationMetricsEasy
• Sequential Test (mSPRT-Style Always-Valid p-Value)StatisticsHard
• Shadow / Canary Rollback DecisionMLSystemDesignAgenticAIMedium
• Softmax from scratchNeuralNetworksEasy
• Thompson Sampling (Bernoulli-Beta)ProbabilityStatisticsAgenticAIMedium
• Two-Sample Test + p-valueStatisticsMedium
• UCB1 multi-armed banditAgenticAIEasy
• Wilson Confidence Interval (Proportion)StatisticsEasy
• WordPiece Tokenization (Greedy Longest-Match)NaturalLanguageProcessingMLSystemDesignMedium