ML Interview Prep

Repository for Machine Learning Engineer (MLE) interview preparation materials and from-scratch algorithm implementations. Additional notebooks and resources will be added here.

📌 Current Status

Available Notebooks

• MLE_Coding_Interview_Guide.ipynb: From-scratch implementations of core MLE coding interview algorithms with mathematical formulations, interview tips, and verification tests:

  1. Linear Regression (Vectorized Gradient Descent & Normal Equation)
  2. Logistic Regression (Vectorized Gradient Descent & Numerically Stable Sigmoid)
  3. K-Means Clustering (Random Initialization, Vectorized Broadcasting & List Comprehension)
  4. K-Nearest Neighbors (KNN) (Unified Pairwise Distance Matrix & $\mathcal{O}(N)$ Top-K)
  5. 2-Layer MLP with Backprop (Manual Multivariate Chain Rule in pure NumPy)
  6. Layer Normalization (NumPy & PyTorch implementations)
  7. Multi-Head Self-Attention (Scaled Dot-Product, Causal Masking & Projections in PyTorch)
  8. Low-Rank Adaptation (LoRA) (PEFT module with initialization trap handling & weight merging)
  9. Byte-Pair Encoding (BPE) (Subword tokenizer with pair-frequency statistics and greedy merging)

• ML_LeetCode_Interview_Guide.ipynb: LeetCode-style algorithmic questions intersecting with Machine Learning, Probability, Retrieval, and Computer Vision in pure Python (no NumPy/PyTorch, standard library only):

  1. Weighted Random Sampling (Prefix Sums + Binary Search / LC 528)
  2. Reservoir Sampling (Uniform Streaming with $\mathcal{O}(1)$ Memory / LC 382 & LC 398)
  3. Sort Transformed Array (Convex Quadratic Parabola & Two Pointers / LC 360)
  4. Find Minimum of a Convex Function (Ternary Search & Divide-and-Conquer)
  5. Expected Calibration Error (ECE) (Bucket Binning & Reliability Gaps)
  6. Exact ROC-AUC with Tie Handling (Wilcoxon Rank-Sum Statistic & Fractional Ties)
  7. Precision-Recall Curve & Average Precision (Sweep-Line Threshold Integration)
  8. Dot Product of Two Sparse Vectors (Two Pointers & Binary Search for Imbalance / LC 1570)
  9. Top-K Cosine Similarity Retrieval (Min-Heap with heapq in $\mathcal{O}(N \log K)$)
  10. TF-IDF & Inverted Index Document Retrieval (Lexical Search & Smoothed IDF)
  11. Streaming Running Median (Dual Heaps in $\mathcal{O}(\log N)$ / LC 295)
  12. Edit Distance & Word Error Rate (WER) (1D Space-Optimized DP / LC 72)
  13. Image Rotation by Arbitrary Angle (Inverse Coordinate Mapping & Bilinear Interpolation)

• AI_Assisted_Coding_Interview_Guide.ipynb: Practical implementation and AI-assisted coding patterns (asked in companies like Shopify, LinkedIn, Meta, and similar tech organizations, with many other companies adopting comparable pair-programming or progressive implementation formats) featuring stateful simulation, robot navigation, game state machines, and progressive class design in pure Python (standard library only):

  1. Mars Rover / Grid Robot Command Engine (Cardinal directions, ray-step obstacle collision avoidance, jump distance, and undo action history)
  2. Blind-Grid Robot Room Cleaner (Sensory API exploration, virtual coordinate tracking, and blind DFS backtracking with physical orientation restoration)
  3. Cleaning Robot with Cycle Detection (Deterministic matrix navigation, obstacle/boundary reflection, and state-space cycle detection)
  4. Snake Game Engine with Growth & Collision (deque body representation, sequential food ingestion, and tail-vacation vs. head collision order)
  5. Connect Four / Tic-Tac-Toe with $\mathcal{O}(1)$ Win Checking (Dynamic counter vectors for rows, columns, and diagonals with $\mathcal{O}(1)$ win evaluation)
  6. Battleship Fleet Coordinator (Grid coordinate validation, multi-deck ship geometry, shot hit/miss tracking, and fleet sinking coordinator)
  7. In-Memory Key-Field Database with TTL & Backup/Restore (Progressive 4-level architecture: CRUD, prefix scan, temporal TTL expiration, and remaining TTL restore)
  8. Cloud Storage Manager with User Quotas & Auto-Eviction (File indexing, user quota enforcement, largest-first tie-broken auto-eviction, and file compression)
  9. Text Editor with Undo/Redo & Find/Replace (Cursor navigation, buffer slicing, regex pattern replace count, and transactional command dual-stack undo/redo)
  10. Multi-Account Financial Ledger with Atomic Rollbacks (Double-entry transaction ledger, point-in-time binary search balance queries, and atomic transfer rollback)
  11. Sliding Window Log & Token Bucket Rate Limiter (Per-client timestamp window log pruning and continuous fractional refill token bucket)
  12. DAG Task Execution Engine with Dependency Resolution (Directed acyclic graph, Kahn's topological sort, circular cycle detection, and task output piping)

🚀 Upcoming Additions

More topic-specific notebooks, ML system design walkthroughs, and practice problems will be added as preparation continues.