Every product has its origin story. DeepExtension is no exception.
When we wrote the first line of code, we weren’t even thinking about building a “standard product.” In fact, we had our own needs but couldn’t find existing tools to fulfill them.
We wanted to fine-tune models, to understand the principles behind fine-tuning, and to see whether tuning really made them “smarter.” At the time, we were a small team with limited resources: tight budget, limited manpower, constrained computing power — and most importantly, limited time. We explored many tools and tried various platforms, but nothing quite fit.
So we built our own. A platform that met our needs and could eventually scale. That’s how DeepExtension was born.
Inspiration: Aristotle in Machine Version
What inspired us most on this journey was a quote from Steve Jobs in a 1983 speech:
“The problem was you can’t ask Aristotle a question…
If we really can come up with these machines…
when the next Aristotle comes around, maybe if he carries around one of these machines with him his whole life,
after the death we can ask this machine.”
This quote deeply moved us. We may never be able to ask Aristotle a question again. But if he could train a machine himself — embedding his thoughts, principles, and methods — then even after he’s gone, we could still “ask him.” That became our founding vision for DeepExtension:
“Aristotle trains Aristotle — in machine version.”
From Expert to AI Embodiment
We believe AGI is coming, but AGI will never be Aristotle, Gaudí, or Buffett.
However, we can build a system that lets these people “teach” AI themselves — to become their embodiment in the silicon world.
That’s what DeepExtension is.
It’s not just an “intelligent platform.” It’s a toolbox that allows experts to train AI to become their digital extension. It empowers professionals — in biotech, material science, industrial design, law, medicine — to fine-tune, evaluate, and deploy AI using their own language, experience, and judgment.
This is not replacement — it’s knowledge transfer: from brain to machine, from carbon-based to silicon-based life.
Why We Built DeepExtension
We want it to be:
- A platform where experts who can’t code can run the entire training workflow
- A collaborative space led by domain experts, supported by engineers
- A scalable system driven by expert input, with AI engineers providing support as needed
- A full-cycle toolchain from prompting to training, testing to deployment
- Most importantly, a container where human knowledge can persist, accumulate, and regenerate
DeepExtension isn’t just a product — it’s our methodology for doing AI. A small answer to the big question of the future.
Final Words
If you’re an industry expert,
if you’ve missed an AI opportunity because you don’t write Python,
if your mind holds a logic deeper than any model…
Then we invite you to use DeepExtension —
to train your “self” — the machine version of you.