AI Engineering
for Software Engineers
AI Engineering
for Software Engineers
AI Engineering
for
Software Engineers
Understand how AI works. Then build and ship systems you can trust.
Understand how AI works. Then build and ship systems you can trust.
with Alex Razvant · Senior Software Engineer, AI
with Alex Razvant · Senior Software Engineer, AI
with Alex Razvant · Senior Software Engineer, AI
AI changes the work. It doesn’t remove the need for engineers.
But it will change what good engineers need to know.
AI changes the work. It doesn’t remove the need for engineers.
But it will change what good engineers need to know.
How do you stay useful as an engineer?
You go deeper.
How do you stay useful as an engineer?
You go deeper.
How do you stay useful as an engineer?
You go deeper.
Writing code was never the whole job. The real work was designing, building, and owning the system around it. As AI finds its way into more products, it helps to know what’s happening underneath—not just how to call a model.
Writing code was never the whole job. The real work was designing, building, and owning the system around it. As AI finds its way into more products, it helps to know what’s happening underneath—not just how to call a model.
Writing code was never the whole job. The real work was designing, building, and owning the system around it. As AI finds its way into more products, it helps to know what’s happening underneath—not just how to call a model.
You might be using a coding agent, adding AI to a product, or stitching together vision, search, and agents. The tools will keep changing. The underlying engineering problems won’t.
You might be using a coding agent, adding AI to a product, or stitching together vision, search, and agents. The tools will keep changing. The underlying engineering problems won’t.
You might be using a coding agent, adding AI to a product, or stitching together vision, search, and agents. The tools will keep changing. The underlying engineering problems won’t.
That’s what The AI Merge is for.
A deeper grasp of how AI works.
The practical skill to build with it.
And the judgment to know when it belongs in a product.
That’s what The AI Merge is for.
A deeper grasp of how AI works.
The practical skill to build with it.
And the judgment to know when it belongs in a product.
That’s what The AI Merge is for.
A deeper grasp of how AI works.
The practical skill to build with it.
And the judgment to know when it belongs in a product.
THE AI ENGINEERING MAP
The layers between a model and a working product.
Prompts are only the beginning. Learn how models behave, how to evaluate them, and how to run the whole thing once real users show up.
01
AI Foundations
Tokens, embeddings, attention, inference, fine-tuning, and the ways models go wrong.
01
AI Foundations
Tokens, embeddings, attention, inference, fine-tuning, and the ways models go wrong.
02
AI Systems
RAG, agents, vision, speech, and memory—and how they fit together in a product.
02
AI Systems
RAG, agents, vision, speech, and memory—and how they fit together in a product.
03
Deployment & Reliability
Serving, APIs, queues, containers, releases, and the unglamorous work of keeping a system running.
03
Deployment & Reliability
Serving, APIs, queues, containers, releases, and the unglamorous work of keeping a system running.
04
Evaluation & Monitoring
How to test quality, spot regressions, compare versions, and see what a model does after it ships.
04
Evaluation & Monitoring
How to test quality, spot regressions, compare versions, and see what a model does after it ships.
05
Compute & Optimizations
GPUs, latency, batching, memory, throughput, quantization, KV cache, and the trade-offs behind the bill.
05
Compute & Optimizations
GPUs, latency, batching, memory, throughput, quantization, KV cache, and the trade-offs behind the bill.
06
AI-assisted Engineering
How to use coding agents like Claude, Codex, Pi, or OpenCode without handing over the steering wheel.
06
AI-assisted Engineering
How to use coding agents like Claude, Codex, Pi, or OpenCode without handing over the steering wheel.
Start with the map. Then build something.
The AI Atlas
A 220+ page visual guide to how modern AI systems fit together.
The AI Atlas is a map of the AI stack, from hardware and training to inference and deployment. It’s for engineers who want to see how the pieces fit.
What you'll learn:
AI Hardware, GPUs, AI Accelerators
Foundations, training, and inference
Agents, applications, and AI systems
What you'll get:
220+ Visual Slides
190+ technical diagrams and system maps
Animated explanations for complex concepts
A map of the AI stack, from hardware to deployment.
The AI Atlas
A 220+ slide visual guide to the AI stack.
See the whole system: hardware, models, training, inference, agents, applications, deployment, and the choices between them.
What you'll learn:
AI Hardware, GPUs, AI Accelerators
Foundations, training, and inference
Agents, applications, and AI systems
INSIDE THE AI ATLAS
INSIDE THE AI ATLAS
A map of what’s happening underneath the AI.
Follow the chain from hardware and training to inference, deployment, and monitoring. It’s for engineers who want to see what makes an AI system hold up in the real world.
See how hardware, models, inference and deployment connect.
Built for software engineers who want the system-level understanding behind AI.
Follow the chain from hardware to deployment. See what makes an AI system hold up in the real world.
220+
Visual slides
Visual explanations you can come back to when the stack gets confusing.
220+
Visual slides
Visual explanations you can come back to when the stack gets confusing.
190+
System Maps
See how hardware, models, training, inference, agents, and apps fit together.
190+
System Maps
See how hardware, models, training, inference, agents, and apps fit together.
4
Core AI Layers
The four layers underneath every AI product: hardware, training, inference, and applications.
4
Core AI Layers
The four layers underneath every AI product: hardware, training, inference, and applications.
∞
Mental Models
Most new AI ideas still come back to the same core concepts.
∞
Mental Models
Most new AI ideas still come back to the same core concepts.
SLIDE EXAMPLES
A few close-ups from inside The AI Atlas.
Drag to explore the map.
Want the whole map?
Join the list for early access to The AI Atlas.
Want the whole map?
Join the list for early access to The AI Atlas.
Want the whole map?
Join the list for early access to The AI Atlas.
I’ll let you know when the Atlas is ready.
Recent notes on building AI systems
Latest
AI Engineering
field notes
Recent notes
on building
AI systems
WHY THIS EXISTS
Less hype.
More engineering.
Each guide and course starts with the whole system, not the latest trick.
BEFORE
AI Summaries
THE AI MERGE
Technical Deep Dives
BEFORE
Prompt lists and demo wrappers
THE AI MERGE
Understanding AI Systems
BEFORE
Isolated tutorials
THE AI MERGE
Fundamentals, engineering and applications
WHY THIS EXISTS
Less hype.
More engineering.
Each guide and course starts with the whole system, not the latest trick.
BEFORE
AI Summaries
THE AI MERGE
Technical Deep Dives
BEFORE
Prompt lists and demo wrappers
THE AI MERGE
Understanding AI Systems
BEFORE
Isolated tutorials
THE AI MERGE
Fundamentals, engineering and applications
Less hype.
More engineering.
Each guide and course starts with the whole system, not the latest trick.
BEFORE
AI Summaries
THE AI MERGE
Technical Deep Dives
BEFORE
Prompt lists and demo wrappers
THE AI MERGE
Understand AI Systems
BEFORE
Isolated tutorials
THE AI MERGE
Fundamentals, engineering and apps
FIELD NOTES FOR BUILDERS
Notes on building AI that are worth your time.
Join 10,000+ engineers. No hype, no prompt lists—just practical notes from the work.
No spam. Just notes on building AI systems.
Unsubscribe any time.
FIELD NOTES FOR BUILDERS
Notes on building AI that are worth your time
Join 10,000+ engineers. No hype.
Just practical notes from the work.
No spam. Just notes on building AI systems.
Unsubscribe any time.
Hey, my name is Alex
I’m Alex Razvant, and I run The AI Merge.
For the past decade, I’ve worked across software, machine learning, deep learning, generative AI, and computer vision.
I’ve gone from publishing computer vision research to building perception systems for autonomous driving, in-store vision, freelance AI projects, and now software and AI systems at Axon.
I make visual guides, courses, field notes, and deep dives about how modern AI systems fit together.
What I want to do is simple: help engineers understand AI well enough to build real systems with it.
I’m Alex Razvant, and I run The AI Merge.
For the past decade, I’ve worked across software, machine learning, deep learning, generative AI, and computer vision.
I’ve gone from publishing computer vision research to building perception systems for autonomous driving, in-store vision, freelance AI projects, and now software and AI systems at Axon.
I make visual guides, courses, field notes, and deep dives about how modern AI systems fit together.
What I want to do is simple: help engineers understand AI well enough to build real systems with it.