Author: Vitaly Rubinovich
Six technical books for building, training, operating, accelerating, improving, and securing AI systems.
Models, prompting, evaluation, retrieval, workflows, tools, agents, and memory.
The mathematics needed to understand how language models are trained and run.
Reproducible runs, distributed training, model serving, storage and retrieval, evaluation, and repeatable workflows.
GPU hardware, CUDA, Triton, profiling, training, inference, and distributed performance.
Policy optimization, preference methods, and reinforcement learning workflows for choosing better outputs.
Threat models, data and model security, agent boundaries, evaluation, operations, and governance.