IM
Stage : Construction automatique de modèles de processus hospitaliers multi- flux par grands modèles de langage (LLM)
Contract type
One-time
Work mode
Remote · Albi, Occitanie, France
Experience
3 years
Job description
A 5-6 month research internship starting in February to design and prototype a pipeline where a single LLM-based process-model construction module is fed by real hospital information system logs (retrospective) and LLM-predicted patient pathways (prospective).
At a glance
- Employment
- Internship
- Work mode
- Remote · Albi, Occitanie, France
- Languages
- French
What you'll do
- Design and prototype a pipeline where one LLM-based automatic process-model construction module is fed by two sources
- Branch A: normalize real hospital information system logs into event logs to produce as-is models of observed processes
- Branch B: serialize LLM-predicted pathways into the same trace format (activity, timestamp, resource) via an adapter to produce to-be models
- Compare as-is and to-be models to detect deviations, anticipate bottlenecks and support explainable decision-making
- Work up to counterfactual simulation: perturb predictions, rebuild the model, observe the impact
What we're looking for
- M2 or final year of an engineering school in computer science, data science, AI or industrial engineering
- Python is essential
- Scientific rigor
- Autonomy
- Taste for working with imperfect real data
- Interest in the hospital domain
Nice to have
- Experience with LLMs (fine-tuning, prompting, API)
- Experience with process mining (pm4py)
Your team
Centre Génie Industriel (CGI), IMT Mines Albi
Project led by Marwa Trabelsi at IMT Mines Albi Centre Génie Industriel (CGI), with IMT partners Télécom Sud Paris (SAMOVAR) and Mines Saint-Etienne (CIS).
Apply
Direct apply at Institut Mines-Télécom
Opens the job posting at Institut Mines-Télécom
About Institut Mines-Télécom
Key facts
- Project acronym HOPAL: Hospital Process Automatic Learning
- Internship lasts 5-6 months starting in February
- Part of an inter-school IMT dynamic around explainable AI for public health
Partners & customers
Telecom Sud ParisMines Saint-EtienneHôpital Le Corbusier de Firminy