Frameworks — Roadmap

The tools you'll actually type into a terminal.

1. Core ML/DL

• PyTorch — tensors, autograd, nn.Module, training loops
• Hugging Face transformers and datasets
• Experiment tracking: Weights & Biases / MLflow

2. LLM / Agent Frameworks

• LangChain — chains, retrievers, agents
• LlamaIndex — data indexing and retrieval for RAG
• An MCP SDK (Python or TypeScript) — building MCP servers/clients
• An agent framework (e.g. LangGraph, CrewAI) for multi-agent orchestration

3. Serving & MLOps / LLMOps

• Model serving: vLLM, TorchServe, Triton
• Containerization & deployment basics (Docker)
• CI/CD for ML: automated retraining and evaluation pipelines
• Observability for LLM apps: tracing prompts/responses, cost tracking