Prove Your Hardest Computer Vision Problem Is Solvable — in Two Weeks, on Your Own Data
A feasibility sprint for founders betting the company on computer vision. You get a working demo on your real footage, a straight go/no-go with the numbers behind it, and a costed path to production. Run by the team behind the open-source Savant video-analytics framework.
From 2 weeks · from $6,000 · you own everything we produce.
You Can't Raise on a Maybe
You've found a computer vision opportunity — but you don't yet know if it's buildable at the accuracy and reliability your product actually needs. Hiring a senior CV team to find out takes months, and a wrong hire costs more than that. A weak prototype burns runway and credibility along with it.
You need a real answer before you commit headcount, a roadmap, or a raise to it.
A Feasibility Sprint That Ends in a Decision, Not a Slide
A Working Demo on Your Data
Not a canned benchmark — your footage, your conditions, your edge cases. You see exactly how it behaves.
The Numbers
Achievable accuracy and latency, and the compute-and-cost envelope needed to hit them — plus, where hardware is a factor, what it takes to run. And the failure modes we found: the ones that matter before you build, not after.
An Explicit Go / No-Go
If it isn't viable at your target cost or accuracy, we tell you, and why. That answer is the point.
A Costed Path to Production
If it's a go: what it takes to ship, where it runs, at what cost and timeline.
From 2 weeks. From $6,000. You keep everything we produce.
From Question to Answer in Four Steps
-
1
Scoping call — free
We pressure-test the problem and tell you honestly whether a sprint makes sense.
-
2
You send data
Representative footage or images from your real operating conditions.
-
3
We run the sprint
Two weeks or more: model selection, training and evaluation on your data, and deployment and latency testing where it matters.
-
4
Readout
Working demo, the report, the go/no-go, and the costed production path.
It Usually Starts Small. Then It Becomes the Product.
A mobile vehicle-recognition platform for municipal parking and compliance enforcement came to us with one question: could plates and vehicles be detected reliably from moving patrol cars, with faces and plates redacted automatically for privacy?
We started with a short feasibility sprint. Today, the system that grew out of it is their production platform — it replaced the legacy computer-vision stack they had previously acquired.
Feasibility sprint → production platform → displaced the incumbent tech they had bought.
The Sprint Is Run by the People Who Ship This to Production
Most feasibility work is done by people who have never had to make it survive contact with real data and real scale. We have. We build Savant, an open-source framework for real-time video analytics — you can read the code. We run computer vision in production across 600+ devices on four continents. The team is ~10 engineers, including PhDs, focused on the hard end of CV: real-time video understanding and custom models for the cases where off-the-shelf ones fail.
So when we green-light a project, it means we can build it — not just prototype it.
Who It's For — and Who It Isn't
This is for you if
- Computer vision is core to your product, and it's genuinely hard — real-time video, a domain where off-the-shelf models fall short, or accuracy and reliability targets generic tools can't meet.
- You're funded and moving toward production, a raise, or a first hire, and you need a real answer first.
- You have, or can capture, representative data from your real conditions.
This probably isn't a fit if
- You need a quick demo with a pretrained model you could wire up in-house in a weekend.
- The problem is a solved, commodity use case with no accuracy, reliability, or latency constraints.
You Don't Have to Build a CV Team Before You Know What You Need
A green light doesn't mean you now have to hire and manage a computer vision team to act on it. We can be that team — carrying the work from sprint through to production and beyond, as your outsourced CV/ML function. Senior computer vision engineering without the six-month hiring cycle, and without the risk of building the wrong team for a problem you've only just de-risked.
Questions Founders Ask Before the Call
Who owns the IP?
You do. You own the models and application code we develop for you. The work builds on our open-source Savant framework, which you're free to keep using.
Is our data and idea confidential?
Yes. We work under NDA, and your footage stays yours.
What if the answer is no?
You still get the report, the numbers, and the reasons — which saves you from spending months and a bad hire finding out the hard way. That's what you're paying for.
Why not just hire a CV engineer?
Eventually you should. But hiring senior CV talent takes months, and you can't scope the right hire until you know what the problem actually demands. The sprint tells you that.
What data do you need?
Representative images or footage from your real operating conditions — the messier and more realistic, the better. We define specifics on the scoping call.
Does this work for our vertical?
We're vertical-agnostic on the vision problem — industrial, smart city, safety, retail, sports, agriculture. What matters is whether it's a hard real-time or custom-model problem.
How much beyond $6,000?
$6,000 / 2 weeks is the floor for a focused, single-question sprint. Broader scope — multiple models, complex deployment — scopes higher. We quote before you commit.
Get a Straight Answer on Your Computer Vision Bet
One free scoping call. If a sprint makes sense, we'll tell you what it costs and what you'll walk away with. If it doesn't, we'll tell you that too.
Free 30-minute call · No obligation · We reply within one business day.