myrecovery is a digital platform for orthopaedic surgical recovery, founded by two UK surgeons, and backed by substantial venture capital. Patients get a mobile app that delivers them timed content specific to their hospital, surgeon and procedure. The app gathers clinical data (including surveys, exercise routine performance, device activity data) and our backend applies data science to deliver clinical insights to care teams.
Orthopaedics and MSK disease is all about movement - gait analysis is the gold standard for quantifying disease progression and outcomes following surgery. Yet, it is rarely performed outside specialised settings such as clinical research and sports medicine as it requires prohibitively expensive equipment.
myrecovery is pioneering the ability to deliver AI-driven movement (gait) analysis of patients undergoing orthopaedic treatment, using consumer hardware. We call this ‘mobile deep MSK analysis’. We allow video of the patient walking, captured with any handheld smartphone, to be processed using state of the art machine learning algorithms to produce an accurate 2D or 3D animated skeleton. The motion of this skeleton can then be analysed for key indicators (range of motion, symmetry, etc.) and results, which are released immediately to the patient and/or clinical team. Currently in the UK, 5% of knee replacement patients return to surgery because their range of motion has not progressed appropriately. By democratising MSK motion analysis, this technology could measure, track and intervene, potentially avoiding costly readmissions.
We have developed a proof-of-concept demo of this capability and are now seeking to productise it. We are seeking a Data Scientist to lead the research and development of the core ML (neural network) component, which will accurately determine 3D human joint positions from mobile video.
The Data Scientist will have the following responsibilities:
- Leading the design and training of a neural network to determine joint positions from video frames
- Leading the programmatic creation of training/validation data sets at scale
- Working with expert gait analysis partners to develop algorithms for analysing gait indicators from joint positions
- Leading the clinical validation of results, working with partner gait labs in the UK
- Working productively as part of a scrum team of 6-8 developers
The following skills/experience are required:
- PhD (awarded or writing up) in Data Science / AI / ML, or equivalent significant commercial experience using image learning
- Expert in theory and practice of ML, including setting up and training of neural networks
- Solid understanding of training networks with image data
- Solid understanding of pros/cons of different network layers (concurrent, pooling, etc)
- Experience of preparing large training and validation sets
- Expert in use of Python for data science (numpy, scipy, pandas, etc)
The following skills/experience are desirable:
- Solid understanding of computer vision techniques
- Experience of training recurrent networks with time-based data sets
- Experience working within multidisciplinary software engineering teams, using collaborative tools such as git
- Experience of using cloud ML resources, including cloud GPUs for training neural networks at scale
Very competitive, depending on experience.
- 25 days paid holiday
- EMI share options scheme
- Bright, modern startup office near Gospel Oak, London
- Developer-spec hardware/software