Jonathan T.

Development of advanced AI pipelines for cardiology and neurosurgery, including classification of 3D voltage maps of the left atrium (feature extraction, t-SNE), classification of arrhythmias from 12-lead ECGs using signal analysis and an InceptionTime deep learning model, and segmentation of the left ventricle.

Design of brain shift compensation algorithms using Coherent Point Drift (CPD) for intraoperative registration, reconstruction of epicardial tissue thickness from point clouds using a signed distance function (SDF), and integration of TAPIR-based video feature tracking.

Projects carried out in collaboration with Johnson & Johnson MedTech.

Led a team of 10 engineers in the development of an AI-based orthodontics software.

Created machine learning, deep learning, and computer vision models in Python, based on research paper consultation. Data preparation, feature extraction, and algorithm implementation.

Production deployment is done via AWS Cloud service, using Sagemaker for training tasks and endpoints, Elastic Container Service (ECS) for task definition and cluster management, Elastic Container Registry for Docker image transfer, and Lambda services for functions. Docker and GIT services are also integrated.

FDA 510(k) submission and CE marking.

Patent submission based on algorithmic innovation.

Secured funding based on AI excellence from the Israeli Innovation Authority.

As an AI engineer, I led the following missions:

1. Definition of the AI workflow and algorithm:

- Formulation and refinement of the AI workflow, with design of adapted algorithms.

2. Construction of coherent datasets:

- Creation of quality datasets, methodical extraction of relevant features.

3. Cutting-edge research:

- Monitoring advancements in 3D segmentation and motion planning models.

4. AI model development:

- Design of machine learning and deep learning models with innovative techniques.

5. Model validation:

- Rigorous validation with k-fold and other methods, adherence to performance criteria.

6. Production deployment:

- Deployment of validated models in a production environment, considering scalability and integration.