Ludovic Jean-Louis

Responsibilities: Provide technical leadership and direction for the Data Science and Generative-AI related topics. Drive best practices for Generative-AI based solutions. Projects Build a conversational agent for psychometrics Goal: Enable HR specialists asking advanced questions on candidates’ psychometrics assessement.

Key accomplishments: Led data science teams across multiple geographies (US and Netherlands) and functions. Designed and implemented AI-driven solutions for conversation segmentation, named-entities detection, and sensitive information identification/de-identification. Trained statistical models optimized for heterogeneous documents (emails, word documents, etc.). Collaborated with product teams to establish data science roadmaps and with People Operations to recruit top talent. Built prototypes of question-answering engines using RAG architecture.

Led a Data Science task force, providing technical direction and expertise. Developed and deployed a classification model for detecting sensitive information in enterprise documents, utilizing binary classification models and cloud-managed platforms. Designed and implemented a RESTful API for training, applying, and publishing statistical models. Built automation pipelines for code compilation, model training, and deployment on Kubernetes.

Contributed in the construction of various tools for exploiting enterprise documents at scale. The projects completed include: Enterprise Search Engine Migration to Apache Solr Key accomplishments: Designed and developed a domain-specific language to query multiple search engines Implemented a document ingestion pipeline for large-scale document indexing Defined index structure and sharding strategy for optimal performance Developed middleware to index and search documents in Apache Solr Content Monitoring Engine for Enterprise Documents Key highlights: Leveraged Lucene to implement a reverse-indexing approach, indexing keywords as documents and using document content as queries Built APIs for content monitoring of emails using Java, Spring, and Swagger

Mitacs Accelerate Postdoctoral grant in partnership with Netmail Inc. Work on various projects in the area of natural language processing applied to email corpus and large scale dataset. The work realised during the postdoc includes building a system for keyword extraction, a system for lexicon expansion and contributing to the development of SemLinker an entity-linking system that participated to the TAC-KBP evaluation. In addition to these projects the postdoc involved supervising 1 student working on de-duplication of documents.

Subject: Supervised and weakly-supervised approaches for complex event extraction and knowledge base population The thesis was in the field of natural language processing and divided into two parts. The first was dedicated to the extraction of complex relations between named entities at a discursive level. The field of application was event extraction in the seismic domain. The second part of the work focused on the knowledge base population task. Precisely, extracting facts concerning named entities to enrich the content of a knowledge base.

Development of various modules for Basel II compliance: risk weighted asset calculation module, exchange rates module, etc. Design and implementation of reports concerning credit risk and market risk. Writing of stored procedures and Data Transformation Service (DTS).