TensorFlow / PyTorch

Works with PyTorch or TensorFlow for training deep learning models. Creates Dataset/DataLoader, builds models via nn.Module or Sequential API. Trains models with basic training loops, logs loss and metrics.

Knows PyTorch basics: tensors, autograd, nn.Module, DataLoader. Uses PyTorch for training simple models and pre-trained LLM inference via Hugging Face Transformers.

Independently develops DL models in PyTorch/TensorFlow for production. Uses PyTorch Lightning or Keras for structuring training code. Applies transfer learning, learning rate scheduling, gradient clipping. Works with GPU training.

Independently develops with PyTorch for LLM: custom datasets, training loops, mixed precision (torch.amp). Uses Hugging Face Accelerate for multi-GPU training and inference.

Designs production DL systems with PyTorch/TensorFlow. Optimizes training through mixed precision, distributed training (DDP/FSDP), gradient checkpointing. Exports models to ONNX/TorchScript for optimized inference.

Designs advanced PyTorch components for LLM: custom attention layers, efficient inference via torch.compile, CUDA graphs. Optimizes training and inference performance at the framework level.

Defines DL framework strategy for the data science team. Establishes training infrastructure standards and best practices. Coordinates GPU resource management and distributed training setup for the team.

Defines PyTorch best practices for the LLM team. Establishes framework usage guidelines, custom extensions, performance optimization. Conducts PyTorch code reviews.

Shapes DL framework strategy at organizational level. Defines investments in GPU/TPU infrastructure. Evaluates emerging frameworks (JAX, MLX) and plans long-term technology decisions for DL development.

Shapes enterprise PyTorch strategy for ML/LLM organizations. Defines approaches to framework management, custom op development, and hardware-specific optimization strategies.