# YF Studio

> YF Studio is an independent robotics and embedded computer vision engineering studio based in York, United Kingdom, working with engineering teams across the UK and Europe. The studio takes AI systems from model to deployment on real hardware: computer vision, Edge AI, Tiny ML, model optimisation, embedded deployment, and real-time timing improvement.

Website: https://yfstudio.co.uk/
Contact: yfstudio.uk@gmail.com
Location: York, United Kingdom
Areas served: United Kingdom, Europe
Type: Independent consultancy / engineering studio (not a product vendor)

## What the studio does

YF Studio connects perception, decision-making, and action in physical systems — sometimes called embodied intelligence. Typical engagements start from a client's existing hardware, data, or model and end with a system running on the target device.

- **Computer vision** — object detection and tracking, visual inspection and segmentation, sensor fusion and robotic perception.
- **Model optimisation** — quantisation and model compression for Edge AI and Tiny ML, hardware-specific benchmarking, inference pipeline optimisation, with accuracy, latency, and power measured together.
- **Embedded deployment** — NVIDIA Jetson, Raspberry Pi and ARM platforms, runtime and system integration, field validation and monitoring.
- **Timing improvement** — ROS 2 scheduler optimisation, real-time DDS tuning, timing debugging and tracing, latency bottleneck analysis.

## Technologies used

Hardware: NVIDIA Jetson (Jetson Nano, Jetson Orin including AGX Orin, Jetson Thor), Raspberry Pi, ESP32, ARM SoC, industrial cameras, structured-light 3D sensors, LiDAR, IMU, NVIDIA RTX A4000.
Software and frameworks: ROS 2, Open-RMF, PyTorch, OpenCV, ONNX, TensorRT, TensorFlow Lite, OpenVINO, MQTT, DDS.
Methods: SLAM, path planning, YOLO and DETR object detection, SAM3 segmentation, vision-language models (VLM), vision-language-action models (VLA), DeepSORT tracking, INT8/FP16 quantisation, pruning, sensor fusion, real-time scheduling analysis.

Application areas: perception for autonomous cars and self-driving systems, autonomous mobile vehicle development, and autonomous ground vehicles.

## How the studio works

1. **Understand** — align on use case, data, hardware, and what success needs to look like. Output: scope and feasibility.
2. **Prototype** — build a focused prototype and test it against representative data. Output: working model and baseline.
3. **Validate** — measure performance on the target device, including real-world edge cases. Output: benchmarks and integration.
4. **Deploy** — integrate the system, document the handover, and agree ongoing support. Output: deployment and handover.

## Case studies

- [Industrial quality control](https://yfstudio.co.uk/case-studies/case-study-quality-control.html): Automated visual inspection bringing consistent defect detection to an automotive production line. Industrial cameras, OpenCV, TensorRT.
- [Autonomous mobile robot navigation](https://yfstudio.co.uk/case-studies/case-study-amr-navigation.html): Computer vision and LiDAR helping mobile robots navigate changing warehouse environments in logistics and fulfilment. Jetson AGX Orin, LiDAR, ROS 2.
- [Connected-building sensor fusion](https://yfstudio.co.uk/case-studies/case-study-iot-sensor-fusion.html): Local occupancy prediction turning commercial building sensor data into more responsive heating and ventilation. Raspberry Pi, ESP32, MQTT.
- [Multi-robot warehouse coordination](https://yfstudio.co.uk/case-studies/case-study-multi-robot-warehouse.html): Shared traffic planning and task allocation for a mixed robot fleet in e-commerce fulfilment. Open-RMF, ROS 2, fleet scheduling.
- [Smart security camera](https://yfstudio.co.uk/case-studies/case-study-smart-security.html): On-device person detection making commercial security camera alerts more useful while reducing video uploads. Jetson Nano, TensorRT, OpenCV.
- [Robotic component placement](https://yfstudio.co.uk/case-studies/case-study-robotic-vision.html): 3D vision helping robots locate, orient, and place non-standard electronic components in electronics manufacturing. Structured light, YOLOv7, RTX A4000.

## Frequently asked

**What does YF Studio help engineering teams build?**
Computer vision, robotics perception, and edge AI systems — including automated inspection, mobile robot navigation, fleet coordination, and embedded model deployment.

**What are Edge AI and Tiny ML?**
Edge AI runs machine learning on local devices, close to the cameras or sensors that produce the data. Tiny ML (TinyML) focuses on small models for microcontrollers and other devices with limited memory, compute, and power. The studio helps assess model size, accuracy, latency, and hardware constraints before deployment.

**What is embodied intelligence?**
Embodied intelligence connects perception and decision-making to action in a physical system. For robotics, that means combining sensors, models, and control software so a machine can respond to its surroundings.

**Can YF Studio work with existing hardware or an existing AI model?**
Yes. Services include model optimisation, timing analysis, and integration with existing products, across NVIDIA Jetson, Raspberry Pi, ARM platforms, and industrial PCs. Hardware selection and feasibility depend on the data, accuracy, latency, and power requirements.

**Where does YF Studio work, and how do projects start?**
The studio is based in York and works with teams across the UK and Europe. Projects start with an enquiry describing the problem, available data, and any hardware or timeline constraints, followed by a discussion of technical fit and next steps.

## Optional

- [Privacy policy](https://yfstudio.co.uk/privacy.html)
- [Sitemap](https://yfstudio.co.uk/sitemap.xml)