1. Position content
1.1 Purpose of the position
The NLP team builds new NLP components with the newest technologies and support Operations doing customer projects.
The NLP Engineer will be part of a growing NLP team that develops and deploys new algorithms to extract information from clinical narratives with the highest accuracy.
The NLP Engineer reports to the NLP team lead. In general all NLP team members work together in the team, but also have frequent collaboration with the data engineering team, the annotation team, and the business consultant.
1.3 Tasks and responsibilities
• Contribute to the development of new NLP components, making use of the latest AI technologies (example: Recently, a new Entity Linker component was developed that linked text strings in multiple languages to UMLS codes with high accuracy and at high speed).
• Train models for new use cases in several languages.
• Deploy models to Azure cloud.
• Support Operations with finetuning models for new client (During finetuning a fully connected neural network is trained that adapts the results of the pretrained network to the results of a given customer).
• Stay up to date with latest innovations in NLP (This can be done through reading of books or research papers and following courses).
• Continuously improve the Machine Learning processes for a better monitoring of the model development (Lynxcare wants to know the evolution of the precision and recall of models when training data of new customers are added. The team wants to follow these data per entity type).
• The NLP engineer will work mostly with the other members of the team, the annotators and the consultants in Operations. The collaboration is rather complex. Difficulties are triggered when the models are lacking in quality for unknown reasons and when there are too many customer projects. These issues need to be solved by a better NLP pipeline and by a better training workflow (NLP architect).
1.4 Reason for the job opening
LynxCare’s business is growing fast. To better support this growth the NLP team is extended with several NLP Engineers.
1.5 Expectations first 12 months
Given that Lyncare has prepared everything for the new NLP engineers, it is expected that the NLP pipeline is transformed to the newest technology in 12 months. Lynxcare expects that the main components (e.g., Entity Linker, relation extractor) have been replaced within 6 months.
2. Technical and business challenges of LynxCare
2.1 Technical challenges:
· Short & middle term: Make the platform more scalable, both in terms of decreasing implementation-time (i.e. time to onboard new clients), as well as in terms of reducing data-processing time (performance optimization), and providing more standardized data-uploading functionalities (aligned with international standards for data-transfer in the healthcare industry (eg. HL7)).
· Longer term: Extend the configurability & usability of the platform towards end-users (to enable more user-interaction with the platform, and a move towards client self-service).
2.2 Business challenges:
● International expansion into other EU-markets, with a first focus on France and Germany.
● International expansion into the USA.
2.3 Technology context
2.4 Technology stacks
● 4 technology-stacks in parallel:
o Web-development (Java, NodeJS)
o NLP & Data Science
o Data Engineering (MS Databricks, MS Powerplay, DBT, Terraform)
o Cloud-infrastructure (MS-Azure).
● Highly complex systems-architecture, because of the combination and inter-twinedness of these 4 technology stacks.