Company Focus
DeepMind Interview Problems (24)
Problems reported from DeepMind data science interviews
| Status | Title | Difficulty | |||
|---|---|---|---|---|---|
| Purpose of Positional Encoding Pro | Easy | ||||
| Vanishing Gradient Problem Pro | Easy | ||||
| Adam Optimizer Internals Pro | Medium | ||||
| DPO vs RLHF Pro | Medium | ||||
| Emergent Abilities in LLMs Pro | Medium | ||||
| Unlock all 24+ problems | |||||
| He Initialization for ReLU Pro | Medium | ||||
| Multi-Head Attention Pro | Medium | ||||
| Positional Encoding in Transformers Pro | Medium | ||||
| RLHF Training Pipeline Pro | Medium | ||||
| Self-Attention Mechanism Pro | Medium | ||||
| Unlock all 24+ problems | |||||
| Xavier Weight Initialization Pro | Medium | ||||
| AdaGrad Limitations Pro | Hard | ||||
| AI Alignment Challenges Pro | Hard | ||||
| Constitutional AI Pro | Hard | ||||
| Decision Tree Stump Pro | Hard | ||||
| Unlock all 24+ problems | |||||
| Exploding Gradients and Gradient Clipping Pro | Hard | ||||
| GAN Training Dynamics Pro | Hard | ||||
| K-Means Clustering Pro | Hard | ||||
| Mixture of Experts Architecture Pro | Hard | ||||
| Neural Network Forward Pass Pro | Hard | ||||
| Unlock all 24+ problems | |||||
| Scaling Laws and Chinchilla Pro | Hard | ||||
| Speculative Decoding Pro | Hard | ||||
| Transformer Architecture Deep Dive Pro | Hard | ||||
| Variational Autoencoder Latent Space Pro | Hard | ||||