Online Workshop
Online Workshop Every Week

Join our free weekly interactive learning sessions.

Master AI/ML with instant feedback and personalized learning

"Cogito, ergo sum" (I think, therefore I am)

— René Descartes

René Descartes
Free Problems
DeepSeek-V3 Reasoning Techniques
This problem set tests understanding of DeepSeek-V3's reasoning-related innovations, including Mixture-of-Experts (MoE), Multi-Head Latent Attention (MLA), FP8 precision training, Multi-Token Prediction (MTP), DualPipe parallelism, and efficiency metrics such as reasoning tokens vs output tokens and active parameters per token. It contains conceptual, numerical, mathematical, and coding tasks. Each question is accompanied by detailed explanations.
56 pts Medium 91 mixture-of-experts deep learning large language models +7
Advanced Transformer Architectures
This problem set focuses on **advanced Transformer architectures** and covers topics such as: - Rotary Position Embeddings (RoPE) - Multi-Linear / Multi-Query Attention (MLA) - Mixture of Experts (MoE) - Training issues (gradient, scaling, distributed) - Inference optimization (KV cache, batching, parallelization) The questions are a mix of multiple-choice, math, and code. You are expected to have strong familiarity with both the theoretical and practical aspects of Transformers. **Environment:** - Python 3.10+ - PyTorch 2.0+
52 pts Hard 92 transformer architectures positional embeddings rotary embeddings +7
Distributed Training (Data, Tensor, Pipeline Parallel)
This problem set focuses on **distributed training strategies** used in large-scale deep learning: - **Data Parallelism** (gradient synchronization, AllReduce) - **Tensor Parallelism** (splitting matrices across GPUs) - **Pipeline Parallelism** (layer partitioning, bubble efficiency)
43 pts Medium 94 distributed training machine learning data parallelism +7
LLM Agents and Reasoning
This problem set covers the fundamentals of Large Language Model (LLM) Agents, a crucial topic in modern AI systems. You'll learn about agent architectures, reasoning patterns, tool calling mechanisms, and practical implementation strategies. The problems progress from basic concepts to advanced implementation, testing your understanding of: - Agent fundamentals and architectures - ReAct (Reasoning + Acting) pattern - Tool calling and function execution - Error handling and retry mechanisms - Agent planning and decision making - Multi-agent systems and coordination Each question is designed to test both theoretical understanding and practical implementation skills. Pay attention to the detailed explanations provided - they contain valuable insights for your AI/ML education journey.
44 pts Medium 98 large language models ai agents external tools +7
Quantization for Efficient Inference
This problem set focuses on **quantization techniques** for efficient deep learning inference. Topics include: - Post-training quantization (PTQ) - Quantization-aware training (QAT) - Symmetric and asymmetric quantization - INT8, INT4, mixed precision - GPTQ, SmoothQuant, QLoRA - KV cache quantization You will answer **20 questions** mixing multiple-choice, math, and code.
55 pts Medium 90 quantization neural networks model compression +7
NumPy and PyTorch Basics
This problem set focuses on testing your understanding of **NumPy** and **PyTorch** functions, as well as how they interact. You will solve 20 multiple-choice questions, starting with basic NumPy operations and moving towards more advanced PyTorch usage and conversions between the two frameworks. Each question includes an explanation for deeper understanding.
20 pts Medium 99 numpy array-shape basic-operations +7
Premium Problems
Python I/O and Data Pipeline Assessment - Part 4
20 questions focused on PyTorch Dataset/DataLoader design: map/iterable datasets, transforms, custom collate/padding, worker seeding/sharding, num_workers/pin_memory/prefetch_factor, caching, memmap/shared memory, batching by size, profiling, and performance tuning.
10.00 60 pts Medium 98 torch.utils.data.dataset pytorch dataset +7
Chapter 02 - Numeric Python
This problem set covers key concepts from Chapter 2: Vectors, Matrices, and Multidimensional Arrays. The problems test understanding of NumPy array fundamentals, including array creation, indexing, slicing, operations, and vectorized computing. Each question is designed to reinforce the core concepts presented in the chapter.
5.00 26 pts Medium 97 numpy-arrays array-attributes shape +7
USAAIO 2025 R1P3 - Logistic Regression Implementation
This problem focuses on implementing logistic regression from scratch using the Titanic dataset. You will work through data pre-processing, mathematical derivations, and implement both gradient descent and Newton's method for logistic regression. The dataset contains passenger information from the Titanic, and your goal is to predict survival based on various features.
10.00 48 pts Easy 93 data-loading pandas data-exploration +7
USAAIO 2025 R1P2 - Basics of Neural Network - From Linear Regression to DNN Training
This problem is about the basics of neural network. Each part has its particular purpose to intentionally test you something. Do not attempt to find a shortcut to circumvent the rule. And all coding tasks shall run on CPUs, **not GPUs**.
10.00 36 pts Easy 96 learning-rate-scheduler pytorch optimization +12
USAAIO 2025 R1P1 - Fibonacci Matrix Form
Let us consider the following sequence: $$ F_n = F_{n-1} + F_{n-2},\ \forall\ n \ge 2. $$
8.00 27 pts Medium 96 fibonacci sequence linear algebra matrix form +7
IAIO 2024 Part 2 - Machine Learning Algorithms and Deep Learning
This problem covers the remaining categories of the 2024 International Artificial Intelligence Olympiad (IAIO), focusing on machine learning algorithms and deep learning. You'll work through practical implementations of k-means clustering, deep learning architectures, and advanced machine learning theory including kernel methods and the Perceptron algorithm. The problems cover: - K-means clustering algorithm implementation and convergence - Deep learning architectures (DALL-E, Transformers) - Perceptron algorithm and kernel methods - Mathematical proofs and theoretical analysis - Parameter counting and computational complexity
10.00 44 pts Hard 99 k-means clustering euclidean distance machine learning +7

Knowledge Graphs

USA AI Olympiad

Explore competitive programming and AI contest preparation concepts

View Full Graph
Grade 5 Math

Discover elementary mathematics concepts and learning paths

View Full Graph

Featured PDFs

View All PDFs
Cover of System Design Interview: An Insider's Guide Volume 2
System Design Interview: An Insider's Guide Volume 2
116 questions 348 pts
Cover of System Design Interview: An Insider's Guide
System Design Interview: An Insider's Guide
108 questions 317 pts
Cover of UNICALLI: A UNIFIED DIFFUSION FRAMEWORK FOR COLUMN-LEVEL GENERATION AND RECOGNITION OF CHINESE CALLIGRAPHY
UNICALLI: A UNIFIED DIFFUSION FRAMEWORK FOR COLUMN-LEVEL GENERATION AND RECOGNITION OF CHINESE CALLIGRAPHY
10 questions 38 pts
Cover of The Principles of Deep Learning Theory
The Principles of Deep Learning Theory
107 questions 418 pts

Featured Books

View All Books
Cover of Acing the System Design Interview
Acing the System Design Interview
153 questions 456 pts
Cover of Numerical Python: Scientific Computing and Data Science Applications with Numpy, SciPy and Matplotlib
Numerical Python: Scientific Computing and Data Science Applications with Numpy, SciPy and Matplotlib
190 questions 543 pts
Cover of Hands-On Machine Learning with Scikit-Learn and PyTorch
Hands-On Machine Learning with Scikit-Learn and PyTorch
200 questions 554 pts
Cover of Deep Reinforcement Learning Hands-On - Third Edition
Deep Reinforcement Learning Hands-On - Third Edition
222 questions 720 pts

Featured Videos

View All Videos
Cover of Flow-Matching vs Diffusion Models explained side by side
Flow-Matching vs Diffusion Models explained side by side
10 questions 29 pts
Cover of Attention in transformers, step-by-step | Deep Learning Chapter 6
Attention in transformers, step-by-step | Deep Learning Chapter 6
10 questions 30 pts
Cover of Knowledge Distillation: How LLMs train each other
Knowledge Distillation: How LLMs train each other
10 questions 27 pts
Cover of Diffusion Model
Diffusion Model
10 questions 32 pts

Popular Topics

machine learning 56 deep learning 40 neural networks 35 reinforcement learning 33 system-design 28 grade5 27 optimization 14 large language models 13 attention mechanisms 13 combinatorics 13 system-architecture 13 natural language processing 12 aime problems 12 Number Sense 12 scalability 11 beginner 10 number theory 10 performance 10 transformers 9 capacity-planning 9
Click on any tag to filter problems by that topic