01
AI does the work
Not a suggestion box. Our products carry the real, repetitive load end to end: grading a stack, designing the part in 3D, preparing the manufacturing file.
An AI company
Where automation stopped, we begin.
We build autonomous systems for work that's too hard, too slow, or too important to leave to today's tools, and every one keeps a human holding the wheel.
How we build
Different products. One principle.
Silvia is an AI company building autonomous systems for work that's too hard, too slow, or too important to leave to today's tools. An AI grader for teachers now, and an AI engineer that designs in 3D, an AI-native operating system, and search built for machines to come. We push the technical edge relentlessly, and every system we ship keeps a human holding the wheel.
02
Humans stay the authority
The person is the referee, never the data-entry clerk. They set the rules, review what matters, and always have the final say.
03
Safety is the default
Conservative about automation by design. Confidence is surfaced, sensitive data is protected, and high-stakes work routes to a human.
04
Depth over breadth
We go deep on one workflow at a time until it genuinely works, instead of bolting forty shallow features onto a chatbot.
From the founder
Why Silvia
I can tell you the exact feeling. Sitting at a desk, doing well, and being quietly certain that doing well at this was not the same as becoming anything.
School hands you a road. It’s paved, lit, patrolled, and everyone you love stands along it clapping. Its promise is a good outcome. Its price is that the outcome is already known. There’s a ceiling built into every safe road, and the ceiling is the point. That’s what safety is: a limit you agree to in exchange for a guarantee.
The other road makes the opposite trade. No pavement, no applause, no promise you arrive at all. Just an honest deal: everything uncapped, including the falling. I don’t think one road is brave and the other cowardly. I think you have to know your own heart rate. Mine only moves on the second road. That seemed like information worth obeying.
Then there’s the car. I’m a car guy, and the car is a Nissan Silvia. If you know, you know. If you don’t: Japan built three icons everyone can name, and then it built the Silvia, the one actual drivers will defend to the death. Perfect balance, a driver’s car to its bones, brilliant in every way except famous. A prince who never got his coronation.
For years I thought I loved that car because it was underrated. The real reason is worse. Some part of me was afraid of becoming it. Excellent and unwitnessed. Full of engineering nobody lined up for. The safe road doesn’t kill what’s in you. It does something quieter. It lets you keep it, unspent, until keeping it becomes the whole story.
So this company is the whole bet, placed in public. That what I have is real. That it belongs to the world and not to a transcript. That proving something to yourself is a legitimate reason to build, and maybe the only one that survives the hard months. That rare is not something you are. It’s something you choose, every day, on purpose, against comfort.
I named the company Silvia because I refuse the ending where the brilliant thing stays a secret.
This is me giving the prince his crown.
— Daniel Lee, Co-founder
The portfolio
Four products, built to the same standard.
Each one does real work in a domain that matters. We're honest about where each stands today.
Building now
In development
Skyline by Silvia
Building now
Grading that ends in your gradebook.
AI grading for teachers. Skyline scores work against your rubric, drafts the feedback, and posts grades to your gradebook. Students can even submit drafts to get feedback before they turn in. You only step in when it matters.
For K-12 teachers and their students
Visit Skyline
✦
Supra by Silvia
In development
An AI engineer that designs in 3D.
An agentic platform for designing and verifying physical products in 3D. State an objective and constraints; Supra plans, builds, tests, and prepares manufacturing-ready output, with human review on safety-critical work.
For beginners who don't know CAD and pro engineers alike
Visit Supra
AI-native OS by Silvia
In development
An operating system built around AI, not apps.
An AI-native operating system driven by intent: you say what you want, and it coordinates the right models and tools to do it, carrying your context between them so many specialist AIs work as one.
For anyone who wants their computer to act, not just open apps
In development
“AI-native OS” is a working name.
Machine Search by Silvia
In development
Search built for machines, not pages.
A search layer designed for AI agents rather than human eyes: structured, high-signal retrieval that machines can query directly to act on, instead of ten blue links to read.
For AI agents and the builders who rely on them
In development
“Machine Search” is a working name.
One company
AI carries the load. People keep the wheel.
Four products, four hard domains, one through-line: do the heavy, repetitive work end to end, then hand the judgment that matters back to a person.
Education
3D engineering
AI-native OS
Machine search
The ambition
A big-dreamer company, built one domain at a time.
We build across very different fields because the same principle holds in all of them: AI can carry the load if a person stays in charge. Wide on purpose, deep by discipline.
Silvia goes deep into work people can't afford to get wrong, and never automates the accountability.
All of this is in progress, not shipped. We'd rather say where we honestly are than dress up a roadmap.
Token Lab YouTube
Beyond our products, we also teach AI in public. It's an independent, editorially honest channel where we test, explain, and build with the latest tools so real people can use them.
Education
Grading and feedback that respects teachers · Skyline
3D engineering
Agentic 3D design through to manufacturing · Supra
AI-native OS
An operating system built around AI, not apps
Machine search
Search built for machines, not pages
Trust & values
The commitments that hold across every product.
These aren't marketing lines. They're constraints we design against from the first commit.
Human in the loop, always
Every Silvia product keeps a person as the final authority. The AI proposes and does the work; the human decides.
Privacy-first
We minimize what we collect and protect what we hold. No training on sensitive user data without explicit consent.
Built for the rules of each field
FERPA- and COPPA-aware for education. Safety review built into engineering. The domain's standards are the spec.
We don't replace people
Our products remove the drudgery so teachers teach and engineers engineer. The work stays human.
Follow along as we build.
We're early, and honest about it. If our approach to AI (real work, human authority, safety by default) resonates, we'd love to stay in touch.