Ghent Metropolitan Area
Since Jan 2025, I started a new endeavour with the startup Rheumafinder to build cloud software as medical device (or SaMD) for the AI detection of rheumatic markers in CT imaging. My main responsibilities are software development and product management but I am also working towards the registration of the device for application in a clinical setting. More information coming soon! From 2016 to 2024, I was working on the development of a SaaS platform to facilitate data analysis for human genetic testing. Software products were MASTR/Alissa Reporter. Initially as a developer and later as a lead, my main focus was the development and verification of products/medical devices (IVD/RUO) for genetic testing. Bioinformatics workflow development was done in Python (Django/JS stack), leveraging AWS cloud. Software development and release went through CI/CD with a release cycle for major software features and patching. CI/CD automation was done through Jenkins/Ansible tower. Efficiency and effectiveness is key to accelerate the road to market. Therefore, I also took lead on innovation by building automated solutions for data assessment and performance evaluation. Working under ISO13485 triggered my interest in RAQA. Herein, I took part in documentation writing, reporting, risk management, but also post-market handling and customer support I developed and led the cloud-based genetic data, master data and gold standard data management system hooked to the platform to improve software application implementation, testing, performance verification, and to ensure legal and regulatory compliance via data governance. I was involved in the management and legal aspects of customer contracts and related data. In 2010, as a bioinformatics researcher, I was fully focused on 'Finding needles in the data haystack'. I worked on a diverse set of life science research projects involving machine learning, data analysis, statistics and data interpretation. Software development was done using Java, Python, R, Matlab, SQL. In 2009, I completed a PhD in applied machine learning within the field of microbiology.
Software application development for hereditary and somatic genetic testing. Driving software-as-medical-device solutions with the Alissa Reporter cloud SAAS framework, from a RUO as well as an IVD angle. Focus on bioinformatics, genetic data analysis and working towards innovative, efficient software solutions for automated data assessment and performance evaluation. Responsible for software lifecycle at bioinformatics applications level: product feature collection and prioritisation, product development, testing and verification, software release (production and beta), patching/hotfixing, 3rd to 2nd level support. People management: 5 team members. Driving bidaily standup, team meetings and one-on-ones. Handling HR aspects, yearly reviews, training, etc. Configuration and maintenance of genetic data management. Improving system implemented in AWS S3, in different software tiers and regions, in close collaboration with Devops team. QA and legal-wise, responsible for data management procedure, setup of customer data sharing MTA template in collaboration with legal team, customer data handling, software customer contract handling (e.g. software beta MTA, data sharing MTA), ensuring traceability and GDPR compliance. Keen interest in RAQA for medical device development and genetic data handling.
Acquisition of Multiplicom by Agilent Technologies. Overlapping and continuation of the role performed at Multiplicom. Conversion of amplication products to the SureSelect hybrid capture products and integration within the MASTR reporter and, later, Alissa reporter software. Conversion implied: embedding of new pipelines, other data, change of system architecture, other bioinformatics approaches, inter-disciplinary collaboration with US teams (on all levels, from R&D to PMO, legal and field). Industrial collaboration with Nvidia on the leverage of GPU accelerated software for secondary NGS analysis.
Bioinformatics and software development for the analysis and interpretation of targeted sequencing data based on amplicon technology. Focus is put on gene panel assays for genetic diseases and oncology, striving towards personalised medicine. Member of the informatics team and collaborating to the development and releasing of a web-based SaaS (software as a service) package for the analysis, visualisation and interpretation of amplicon data and corresponding variant detection. 1. Bioinformatics: NGS data analysis and variant calling in germline and somatic/tumor samples. - Research and assessment of related open-source tools for bioinformatics analysis of amplicon sequencing data - Parameter optimisation and statistics - Statistics and data visualisation - Reference data management and evaluation - Collaboration with other departments for assay verification and validation studies, risk assessment, marketing input, regulatory aspects, service lab studies, etc. - Working towards releasing RUO and CE-IVD products within a software platform 2. Software development: building software pipelines for bioinformatics analysis - Programming of in-house workflows and tools. Languages: Python (general, pandas, numpy, scipy, etc.), YAML, R - Customisation of open-source tools: Java/Python - Automation of reference data evaluation - Participation in commercial software release (unit/functional testing, QA, software lifecycle, ...) 3. Data management - Data governance - Structuring and standardisation of reference and master data - Involved in ISO certification (ISO27001) 4. Business: - Product development under ISO13485. Involved in procedure and report writing, verification and validation, risk management, 2nd level support - People management (entry level) - Training: Business, Quality Assurance
Bioinformatics data analysis and software development, integration and automation. Main topics: 1. computational data analysis and machine learning within the fields of general bioinformatics, systems biology, genomics, transcriptomics, phenomics and next-generation sequencing 2. data integration and automation 3. data interpretation and visualisation 4. bioinformatics software development Project-based work in collaboration with bioinformaticians and wet-lab researchers of internal laboratories Scientific collaboration with external institutes and companies. Groups: Member of group Yves van de Peer (2010-2014) Member of Applied Bioinformatics and Biostatistics (2014-2016), collaboration with prof. Steven Maere and Dirk Inzé
Keywords: Machine learning, Data analysis, Bioinformatics and Microbiology At 04/12/2009, I defended my PhD dissertation 'From learning taxonomies to phylogenetic learning: a computational approach to a FAME-based bacterial species identification'. In this research, I evaluated the application of machine learning techniques to improve bacterial species identification based on fatty acid methyl ester (FAME) content. Gaschromatographic FAME analysis is routinely performed in microbiological laboratoria for phenotypic characterisation. Bacterial identification is commercially exploited by MIDI Inc. (USA). This research was a new stage in FAME analysis at the Laboratory of Microbiology (Ghent University) and focussed on the database resulting from 20 years of FAME analysis.. In a first part, FAME data sets were created for 3 model genera (Bacillus, Paenibacillus and Pseudomonas). An initial data analysis was performed to investigate how the different FAME profiles relate to each other at species level and led to an assessment on the performance of machine learning techniques to distinguish bacterial species within a single genus. In a second part, we tried to distinguish the different Bacillus species using artificial neural networks by different strategies. Subsequently, reseach was lifted to the classification of species of the three chosen genera. We also evaluated support vector machines and random forests. Given the restrictions of FAME analysis for bacterial species identification, (relatively) good results were obtained within each genus. Results clearly improved the identifications by the commercial system. Finally, different independent data sets were identified. In a third part, we developed a method to investigate the restrictions of FAME analysis and to put identifications in a better context. Research was based on the genus Bacillus and tested by the two other genera. In a fourth part, a public FAME database was developed (http://www.fame-bank.net)