Principal Data Scientist (Computer Vision)

About the project

They are a spatial computing platform that employs virtual 3D replicas of physical locations – called Digital Twins – to help customers visualize, collaborate, and work in physical spaces more creatively and efficiently.

They help real property owners, managers, service providers, and end users make good decisions by providing comprehensive location intelligence in a digital twin of a place.

Their collaborative tools help clients leverage this intelligence to drive new efficiencies and deliver smart user experiences while fostering agility and sustainable growth for our customers and communities.

Stage of development: Have developed a basic prototype to include image recognition and object detection (input). As well as some accuracy of the bounding box where they take the pixel coordinates of that bounding box corners. The accuracy of the positions depends heavily on the accuracy of the bounding box. They need NL to help with this and also want it to be self-learning/trainable and easy to use.

Objective

The main problem is that assets in the design have now moved and are unable to track changes in the environment. The client started to work on a solution to solve this. They combined object detection libraries and used data from object detection to geo-locate the object by getting bounding boxes around the objects. They need to build this out and don’t want to allocate internal resources to do this and would rather enlist a provider who could take over and build this better and faster. So, time is another pain point.

Areas of Responsibility

• AI solution architecture design and roadmap planning

• Engineering team leadership and performance management

• Communication with the customer on the development progress

• AI solution technical quality and performance management

Skills

• Python software development

• AWS solution architecture development

• Model development in AWS Sagemaker

• Training computer vision models (CNNs, Transformers) from scratch

• ML model optimization and compression

• Working efficiently with software engineers, data scientists, stakeholders, etc.

• Clear and concise communication, especially of complex technical concepts to non-technical stakeholders.

Knowledge

• Modern computer vision neural network architectures: convolution-based (i.e. YOLO) and transformer-based (i.e. ViT)

• Deep learning frameworks: Tensorflow, PyTorch

• ML model optimization frameworks

Experience

• Building computer vision algorithms hands-on, 5+ years

• Optimizing and deploying computer vision algorithms on edge architectures, 3+ years

• Development with major cloud providers (AWS, Azure, GCP), 3+ years

• Leading machine learning teams, 2+ years

• Experience with the ONNX framework/willing to learn it

Terms & conditions

Allocation: 0.5 FTE

Time zone: preferably Europe

Candidate’s location: preferably Europe

Start date: October 2023