Question 1

What does The Noughty Fox specialize in?

Accepted Answer

We're a mobile computer vision and AR R&D studio. We build on-device detection, tracking, and 3D reconstruction features that run natively on phones — no cloud dependency at the scan step, no external hardware. Think real-time object detection with custom-trained models, body and face tracking, spatial capture, and AR experiences. If it involves a phone camera doing something intelligent, that's our lane.

Question 2

What's the difference between you and a general mobile development agency?

Accepted Answer

Most agencies build screens and API integrations. We build what's behind the camera button — the computer vision pipeline, the ML model running on-device, the ARKit/ARCore session that makes sense of the physical world. Our team has deep experience in areas like CoreML model training and deployment, MediaPipe integration, LiDAR-based spatial capture, and real-time 3D projection. These aren't skills you find on a typical agency bench.

Question 3

Do you only work on iOS?

Accepted Answer

iOS is where most of our production work lives — ARKit, Vision, CoreML, Metal. But we also build cross-platform solutions. We've delivered React Native bridges over native SDKs, and our spatial capture work is expanding to Android. If your product needs to run on both platforms, we'll scope what makes sense technically rather than default to one.

Question 4

How does an engagement typically start?

Accepted Answer

Most engagements start with a focused proof of concept — a small, paid project (typically 3–5 weeks) scoped to your hardest technical question. The goal is simple: prove we can solve the problem before either of us commits to a larger engagement. We carry the risk, you get a working prototype and a clear picture of what a full build looks like.

Question 5

Why a proof of concept instead of jumping straight into a full project?

Accepted Answer

Because computer vision and AR work has high technical uncertainty. Can the model detect your specific objects accurately enough? Does the tracking hold up in real-world lighting? Will it run at 30fps on target devices? A POC answers these questions with real code on real devices, not with slides. It also lets you evaluate how we work — our communication, code quality, and pace — before committing to a longer engagement.

Question 6

What happens after the POC?

Accepted Answer

If the POC proves the concept works, we move to a full development engagement. That can be a fixed-scope project or ongoing team augmentation where our engineers work as an extension of your team. Most of our long-term clients started with a single POC and expanded from there.

Question 7

How big is your team?

Accepted Answer

We're a team of about 14 people, 10 of whom are engineers. Small enough that the founders (CEO and CTO) are directly involved in every project, large enough to staff serious multi-month engagements. We're based in Chișinău, Moldova, and work across European and US time zones.

Question 8

Can you train custom ML models for our specific use case?

Accepted Answer

Yes. We've trained and deployed custom detection models (YOLO-based) for clients who needed to detect things no off-the-shelf model covers — specific retail products, golf balls on grass, radiators in rooms. If you have or can collect a labeled dataset, we'll train, optimize, and deploy a model that runs on-device via CoreML or equivalent.

Question 9

Do you build SDKs that other developers integrate?

Accepted Answer

This is one of our strongest areas. We've built and shipped production SDKs consumed by third-party development teams — including binary distribution via Swift Package Manager, React Native bridges, full API surfaces with authentication, and comprehensive sample applications. We understand what it takes to make an SDK that external developers actually want to integrate.

Question 10

What AR frameworks and tools do you work with?

Accepted Answer

ARKit (world tracking, face tracking, body tracking, RoomPlan), ARCore, Vision framework, CoreML, MediaPipe (pose, face, hand), Metal for GPU rendering, OpenCV, YOLO (v5/v8/v11), and custom ML pipelines. We also work with point cloud processing, Gaussian Splatting reconstruction, and sensor fusion (IMU + camera).

Question 11

Where are you based? What time zones do you cover?

Accepted Answer

Chișinău, Moldova (EET/EEST — GMT+2/+3). We overlap comfortably with all European time zones and have 4–6 hours of overlap with US East Coast. Most of our clients are in Western Europe and Israel.

Question 12

How do you communicate during a project?

Accepted Answer

We adapt to your workflow. Most clients use Slack for daily communication, with weekly or bi-weekly sync calls. Code lives in your repositories (or ours, depending on the setup). We push regularly, do code reviews, and integrate with your CI/CD. The goal is to feel like we're in the same office, not across a continent.

Question 13

What does pricing look like?

Accepted Answer

Our POC engagements are intentionally affordable — designed to prove value, not extract budget. For ongoing work, we typically charge on an hourly or monthly retainer basis, depending on the engagement structure. We'll give you a clear estimate before any work starts, and we don't surprise you with scope creep.

Question 14

Do you sign NDAs?

Accepted Answer

Yes, standard practice. We're comfortable with mutual NDAs and can work within your confidentiality requirements. Several of our reference projects are under NDA, which is why we describe capabilities without always naming the client.

Question 15

How long have you been doing this?

Accepted Answer

The Noughty Fox was founded in 2021, but our founding team has been building mobile AR and computer vision products for 5+ years. Our prior work includes projects that were acquired by Apple. We've shipped production code for clients including major retail platforms and international tech companies.

Question 16

Can you share references or case studies?

Accepted Answer

Some of our work is public (check our Portfolio), and some is under NDA. We're happy to discuss relevant experience on a call and can arrange introductions to past clients when appropriate. Our Clutch profile and Upwork history also reflect our track record.

Question 17

What if the POC shows the problem can't be solved?

Accepted Answer

It happens — and that's exactly the point of the POC. If the technical approach doesn't work within the constraints, you'll know early and at a fraction of the cost of a full project. We'll deliver a detailed report on what we tried, what the limitations are, and what alternative approaches might work. You keep all the code and findings.

Question 18

What kinds of problems have you solved before?

Accepted Answer

A few examples from our production portfolio:  
- *3D spatial capture from a phone* 
— A platform that captures rooms using LiDAR and ARKit, generates linked panorama tours, point clouds, and floor plan inputs. No external hardware, no trained operators. 
- *Body tracking for fitness coaching* 
— On-device pose detection with exercise counting, movement error detection, and real-time feedback using MediaPipe and custom temporal ML models. 
- *Face tracking for aesthetics and dental* 
— AR face sessions with precise landmark alignment, lighting analysis, face segmentation, and guided capture flows for clinical and cosmetic applications.