AI and LLMs learning roadmap

What it can do, how it works, and where it fails

Asking, looking things up, checking facts, and using AI safely

Syntax, data structures, and writing scripts

Linear algebra, calculus, probability, and statistics

Training, generalization, evaluation, and classic algorithms

Neural nets, backpropagation, and attention

Tokenizers, models, datasets, and fine-tuning

Turning single calls into reusable workflows

Retrieval, vector stores, reranking, and evaluation

Tool use, workflow orchestration, and agentic RAG

Test sets, error analysis, and the improvement loop

Quantization, throughput, and latency tradeoffs

Implement attention and the training loop line by line

SFT, DPO, and online reinforcement learning

Data, systems, scaling laws, and distributed training

Judging demand, data, ethics, and project management

Pretraining, post-training, inference, and why models hallucinate

Prompts, RAG, UX, and LLMOps

Test sets, error analysis, and the improvement loop