ML System Design

90 problems

• Find the bottleneck executorMedium
• Format retrieved docs as contextEasy
• MockSearchEngine keyword retrievalEasy
• The RAG run: search → ground → answerMedium
• Aggregate trace eventsEasy
• Run totals from the traceEasy
• BM25 ScoringMedium
• Bradley-Terry Strength MLEHard
• Citation Attribution AssemblyMedium
• Clipped Inverse Propensity WeightingMedium
• Continuous Batching SchedulerHard
• Cross-Encoder Rerank Top-kEasy
• Edge models — magnitude pruningEasy
• Image preprocessing — scale and standardizeMedium
• RGB to grayscale (luminosity)Easy
• Edge models — measure sparsityEasy
• DDM Concept Drift DetectorMedium
• Elo Rating UpdateEasy
• Expected Reciprocal Rank (ERR)Medium
• Exponential Backoff + JitterEasy
• Faithfulness / Groundedness ScoreMedium
• Feature Hashing (Hashing Trick)Medium
• Gradient checkpointing forwardEasy
• Group / Time-Series CV SplitEasy
• Hard Negative Mining for RetrieverMedium
• HNSW Greedy Graph SearchHard
• Post-Training int8 QuantizationMedium
• Team-Draft Interleaving (Online Ranking Eval)Medium
• IVF (Inverted File) SearchMedium
• Jensen-Shannon Divergence (Drift)Medium
• Kendall's TauMedium
• Knowledge-Distillation LossMedium
• Kolmogorov-Smirnov StatisticMedium
• KV-Cache Eviction (Sliding Window)Medium
• Label-Noise Detection (Confident Learning)Hard
• LambdaMART Lambda GradientsHard
• ListNet Listwise Ranking LossMedium
• Little's LawEasy
• Magnitude PruningEasy
• Mann-Whitney U TestMedium
• Sentence Embedding (Masked Mean-Pool)Easy
• Matryoshka Embedding TruncationMedium
• Mean Average Precision (MAP)Medium
• Mean reciprocal rank (MRR)Easy
• MinHash for Near-Duplicate DetectionMedium
• MIPS -> L2 ReductionMedium
• Maximal Marginal Relevance (MMR)Medium
• Multi-Query Fan-Out + DedupEasy
• Multi-Task Score FusionMedium
• Multi-Tool Router (by Embedding)Medium
• Mutual informationMedium
• Negative Sampling (Implicit Feedback)Medium
• Out-of-Fold Target EncodingMedium
• Paged-Attention KV Block AllocationMedium
• Pairwise cosine-similarity matrixEasy
• Pairwise Ranking Loss (BPR / RankNet)Medium
• Parent-Document RetrievalMedium
• Permutation Feature ImportanceMedium
• PII Detection & RedactionMedium
• Point-in-Time Feature JoinMedium
• Population Stability Index (PSI)Medium
• Precision@k and NDCG@kMedium
• Prefix-Cache Hit AccountingEasy
• Product Quantization (PQ) Encode + ADCHard
• Random-Hyperplane LSH (Cosine)Medium
• Recall@k (Retrieval)Easy
• Reciprocal Rank Fusion (RRF)Medium
• Recursive Chunking with OverlapMedium
• Refusal / Confidence GateEasy
• Reservoir samplingMedium
• Run-Level Cost / Budget CeilingMedium
• Sample Ratio Mismatch (SRM) CheckMedium
• In-Batch Sampled Softmax + logQ CorrectionHard
• Semaphore-Bounded Concurrent TasksMedium
• Prepare RNN data — scale, window, reshapeMedium
• Sequence padding & maskingEasy
• Shadow / Canary Rollback DecisionMedium
• Sliding-Window Counter (Streaming Feature)Medium
• Smoothed Target (Mean) EncodingMedium
• Sorted polynomial featuresMedium
• Spearman Rank CorrelationEasy
• Streaming SSE Delta AssemblerEasy
• Stop-Sequence Detection (Streaming)Easy
• Task-DAG Parallel SchedulerHard
• Temporal Train/Val SplitEasy
• TF-IDFMedium
• Token-Bucket Rate LimiterEasy
• Trace Aggregation SummaryEasy
• Two-Tower Top-k RetrievalEasy
• WordPiece Tokenization (Greedy Longest-Match)Medium