MD and Biomedical Informatics Researcher2020 - PRESENTRouen University Hospital
Technologies: Docker, Python, Vue, Natural Language Processing (NLP), BERT, Data Analysis, APIs, Pandas, Big Data, Data Science, API Development, Flask, Web Development, Web Scraping, Jupyter, Data Analytics, TypeScript, REST APIs, SQLAlchemy, NoSQL, YAML, Python 3, Document Parsing, Scraping
- Performed NLP research in healthcare, consuming the Wikimedia API to augment medical thesaurus translations: https://mikaeldusenne.com/wikimesh.
- Conducted research on using BERT NLP embeddings applications on medical documents and BERT on the hospital's clinical data warehouse.
- Researched medical ontologies and data quality improvements by augmenting existing terminologies in the Medical Subject Headings (MeSH) thesaurus by adding translations from Wikipedia.
- Developed the API consumer in Python, the demonstrator in Vue.js, and evaluated the performance and usability of the solution.
Full-stack Developer2021 - 2022BlackPool (Freelance)
Technologies: AWS Lambda, Vercel, Web3.js, APIs, Python, Vue, Docker, Blockchain, GraphQL, Serverless Architecture, Front-end, React, Non-fungible Tokens (NFT), Amazon Web Services (AWS), CSS, HTML, Bootstrap, Responsive Web Design (RWD), Databases, Pandas, Big Data, Data Science, API Integration, API Development, Flask, Web Development, Next.js, Tailwind CSS, Lambda Functions, Data Analytics, TypeScript, REST APIs, NoSQL, YAML, Ajax, CSS Grid, Python 3, Document Parsing, Scraping, Amazon S3 (AWS S3)
- Developed a subdomain of the website related to the company's Sorare fantasy NFT football game. This involves ETL from an API and provides custom data analytics tools: https://sorare.blackpool.finance.
- Queried a GraphQL API to obtain data related to the company using a serverless architecture. Used AWS Lambda and Scheduler to automatically feed the cloud-housed MongoDB database.
- Displayed the raw data using Vue.js and a static website deployed via Vercel. Created custom aggregated reports to provide advanced analytics tools for statistical analysis and exploration to provide accurate insights.
Full-stack Developer2020 - 2021Freelance
Technologies: Docker, Python, Vue, MongoDB, Front-end, Node.js, Amazon Web Services (AWS), CSS, HTML, Bootstrap, Responsive Web Design (RWD), Databases, NumPy, Pandas, API Integration, API Development, Flask, Web Development, Web Scraping, Lambda Functions, Jupyter, Data Analytics, TypeScript, REST APIs, SQLAlchemy, NoSQL, jQuery, YAML, Ajax, CSS Grid, Python 3, Document Parsing, Scraping, Amazon S3 (AWS S3)
- Enabled my client to update his workflow and stop sending and receiving Excel files via email. Developed a complete website for my client and his clients.
- Improved the client's UX and attractivity by replacing Excel-based questionnaires with a web platform featuring secure user authentication and a UI created with HTML with Vue.js.
- Added a data processing pipeline, replacing Excel formulas with documented and debuggable Python scripts.
- Increased productivity by automating processes, reduced errors by eliminating copy-paste, and added new functionalities that had been too hard to implement.
- Improved the ease of use and user confidence by adding a secure payment module that allowed users to make purchases through the platform instead of making manual arrangements with my client.
Medical Resident | Medical Informatics Researcher2019 - 2020Rouen University Hospital
Technologies: Doc2Vec, Docker, Vue, Python, Word2Vec, Natural Language Processing (NLP), Plotly, Front-end, CSS, HTML, Bootstrap, Responsive Web Design (RWD), Databases, NumPy, PyTorch, Web Development, Web Scraping, Jupyter, Data Analytics, REST APIs, NoSQL, jQuery, YAML, Ajax, CSS Grid, Python 3
- Explored the utilisability of an NLP neural network-based embedding technique for medical document classification. Published an article https://www.mikaeldusenne.com/article_ia_sante_export.pdf and presentation: https://www.mikaeldusenne.com/ias/.
- Presented the research during a French national congress on artificial intelligence in healthcare.
- Implemented Doc2Vec to generate an embedding representation of medical documents in the hospital's clinical data warehouse.
- Used the embeddings to perform automatic classification of medical document types with great accuracy, improving the data quality in the clinical data warehouse.
- Created a Vue.js interface to test the model, allow medical doctors to easily classify documents to doctors to evaluate the model's performance, and generate automatic reports.
Medical Resident | Medical Informatics Researcher2018 - 2019Harvard Medical School, Avillach Lab at the Department of Biomedical Informatics
Technologies: Computer Science, Data Analysis, DevOps, Docker, R, Machine Learning, SQL, Slurm Workload Manager, Job Schedulers, Python 3, Document Parsing, Amazon S3 (AWS S3)
- Analyzed a clinical registry to identify comorbidity associations in patients presenting with an orphan disease. Published an article written as a medical thesis: https://dumas.ccsd.cnrs.fr/dumas-03116249.
- Explored machine learning techniques to answer the research question using an SQL database and R for analysis.
- Used the department computer cluster with the Slurm job scheduler to perform parallelized computation.
Back-end Developer2017 - 2019Inserm: National Institute of Health and Medical Research (Freelance)
Technologies: Docker, Python, Vue, Python 3
- Implemented the semantic interoperability module for the C3-Cloud project, a European project aiming to improve medical decisions for aging patients with multiple comorbidities.
- Wrote a REST API in Python (Flask) with an SQLite database, exposing manually curated mappings between several medical terminologies compliant with the FHIR standard. Demonstrator: https://rubis.limics.upmc.fr/c3-cloud/#/.
- Deployed Docker containers with a well-documented and reproducible deployment procedure, enabling easy integration with the other project components.
- Collaborated with international teams developing other modules, provided a well-documented and flexible deployment process, and maintained good communication with the teams.