Cover and press assets
The cover of How to Make Your Model Fast in the sizes you need for a post, a listing or an article, plus a short description you can quote.
You are welcome to use any of these when you write about or share the book. No permission needed; a link back to ai.usamah.me is appreciated.
- Cover, 1600 x 2560 (e-book listings, Amazon and Apple Books proportions)
- Cover, dark version, 1600 x 2560 (dark-mode feeds, or a hard edge on a white page)
- Link card, 1200 x 630 at 2x, 2400 x 1260 (LinkedIn, X, Slack and Open Graph previews)
- Square, 1080 x 1080 (Instagram, LinkedIn and X image posts)
- Portrait, 1080 x 1350 (Instagram 4:5 feed)
- Story, 1080 x 1920 (Instagram and LinkedIn stories)
The picture on the cover comes from the introduction: a robot has 33 milliseconds to see and react, and here is where 36.6 of them went. The stage timings are illustrative, and they add up.
Alternate covers
Seven designs were made for the cover. These are the others, if one of them suits a particular post better:
- Roofline, dark and Roofline, light, 1600 x 2560: the roofline model from Part 1 as the whole cover, with its own link card, square and story
- Textbook, 1600 x 2560: a classic technical-press cover with the roofline drawn as Figure 1
- Stack, Typographic, Die and Field, 1600 x 2560
About the book
How to Make Your Model Fast: A Systems View of Efficient Machine Learning, from Silicon to Agents is a free, web-first book by Usamah Zaheer. Most of what is written about machine learning assumes the model is the interesting part and the machine is a detail. In practice the machine decides what you are allowed to build. The book is about the boundary where a model meets real hardware under a real budget for latency, memory, power and money: how to predict what that boundary will do to you, and what to change first when it does. It runs from the silicon upward, through rooflines, edge accelerators, kernels, compilers, quantisation, pruning and distillation, vision, language models on small machines, robotics, profiling, serving and agents, in fourteen parts.
About the author
Usamah Zaheer is a machine learning software engineer in Cambridge, UK, working on ML compilers, libraries and inference performance for Arm architectures. He was previously a robotics software engineer at Dyson, putting vision models on robot hardware, and before that worked on satellite imagery pipelines at the University of Leicester. He writes at usamah.me.
How to credit
Usamah Zaheer, How to Make Your Model Fast, ai.usamah.me.
Questions, corrections or interview requests: usamahzaheer155 [at] gmail [dot] com.