Etienne Ferrandi

Design of a conversational agent using LangGraph and LangChain for medical kit tracking software Implementation of RAG techniques (hybrid search with Qdrant, LLM Wiki) PDF parsing, OCR, etc. Calling LLM APIs via AWS Bedrock Deployment on AWS via Bedrock Agentcore Development of time-series forecasting models using "encoder-only" foundation models Anomaly detection in tracking data

Utilizing data mining methods to identify unstructured bibliographic data using LLMs deployed on self-hosted GPU VMs via vLLM Structuring this data using RAG (Qdrant) and constraint-based generation (Pydantic) Fine-tuning language models, including mT5 (encoder-decoder) and Llama 3.1 8B (decoder-only) using the Transformers library

Support and management of projects focused on Generative Artificial Intelligence Development of a RAG conversational agent (Albert API) incorporating chunking and vector integration of XML data to assist public officials Extraction of unstructured bibliographic data from the 20th-century archives of the Defense Historical Service using OCR/HTR models Automated generation of XML/EAD encodings for archival collections Project Management Support for the Implementation of Semantic Graph Tools Applied to Research Data for the IRD

Redesign of an intranet for Sources Chrétiennes - Upgrade to the latest version (PHP 8.1) - Migration of the MySQL database Update of the biblindex website - Creation of the user journey (registration, login, forgotten password, etc.) - Definition of a Twig template for sending emails - Configuration of user permissions in the EasyAdmin interface - Definition of profile preferences

Technical translation of the business requirements of researchers in the humanities and social sciences for the SourcEncyMe project - Encoding raw texts in XML/TEI format - Deep learning from scratch (named entity recognition, NER)