Tremplin Recherche · ESIEE Paris

Building NLP systems that read clinical language the way oncologists do.

I'm Clément Longeac, an engineering student working at the intersection of clinical NLP and biomedical AI — currently building hybrid extraction pipelines for French oncology records, on the way to a PhD in AI applied to biomedical research.

DEMNE — Rules → CRF → Transformer → LLM, live schematic

From tutoring math to modeling tumor markers

I'm a fourth-year student at ESIEE Paris, specializing in Data Science & AI, and enrolled in the school's Tremplin Recherche track — a program built for students heading toward doctoral research rather than a standard industry placement.

My work centers on making unstructured clinical text usable for research: extracting the biomarkers, staging information, and treatment details buried in French oncology reports, where general-purpose NLP tools consistently fail on domain vocabulary and document structure.

Alongside my studies I tutor mathematics, physics, and computer science through Complétude, and I'm preparing applications for a CIFRE industrial PhD in biomedical AI at Sorbonne University.

Institution
ESIEE Paris
Program
Data Science & AI (DSIA)
Track
Tremplin Recherche
Next step
PhD, biomedical AI — Sorbonne
Languages
FR · EN · JP · ES

DEMNE / DuraXeLL

A hybrid named-entity recognition pipeline for French oncology clinical text.

Determination of Extraction Methods for Named Entities

Clinical NER · French oncology corpora

Off-the-shelf NER models struggle badly on French clinical text: dense abbreviations, inconsistent formatting across hospital systems, and biomarker terminology that barely appears in general training data. DEMNE tackles this with a staged pipeline rather than a single model — each stage handles what it's actually good at, and only escalates to a heavier model when it has to.

The system is fine-tuned on DrBERT-7GB across three distinct French oncology corpora, and targets seven priority biomarker entity types used in treatment-relevant reporting.

Clinical NLP CRF Transformers DrBERT-7GB French oncology
01

Rule-based pass

Regex and lexicon matching catch high-confidence, low-ambiguity entities cheaply — dates, dosages, standard biomarker abbreviations.

02

CRF layer

A conditional random field models sequential structure the rules miss, using local context to resolve ambiguous spans.

03

Transformer extraction

DrBERT-7BG, fine-tuned across three oncology corpora, handles the entities that need real semantic and domain understanding.

04

LLM arbitration

Remaining conflicts and edge cases are routed to an LLM pass for final disambiguation before output.

Where the work has happened so far

2025 — Present

Research Assistant, Tremplin Recherche

ESIEE Paris

Developing DEMNE, a hybrid clinical NER pipeline for French oncology text, as the core project of the Tremplin Recherche track ahead of PhD applications.

2025

Engineering Intern, GRADES Division

Synchrotron SOLEIL

Developed autopkgtests using OpenCL for AMD GPU compatibility (pyopencl, nabu, ufo-filters, pyvkfft) across Debian/ROCm infrastructure on GitLab Salsa.

2022 — Present

STEM Tutor

Complétude

Teaching mathematics, physics, and computer science to secondary and university students alongside my engineering studies.

Notes from the research

Longer write-ups on the methods, dead ends, and decisions behind the work above.

Draft — coming soon

Why French clinical NER breaks general-purpose models

A walkthrough of where standard NER pipelines fail on French oncology reports, and what that implies for how DEMNE is structured.

Draft — coming soon

Staged extraction vs. a single large model

The case for routing entities through rules, a CRF, and a fine-tuned transformer before ever calling an LLM — and what it costs to get wrong.

Let's talk about clinical NLP, biomedical AI, or a PhD collaboration.