CAREERS AT XPDEEP
Build what machines decide with.
Machines Will Decide. We Decide How. Xpdeep builds control infrastructure for bounded AI authority: natively explainable deep-learning models for intelligent machines and consequential industrial systems, whose decisions can be read, bounded and evolved under control.
01 THE TECHNICAL CHALLENGE
Make a learned model carry decisions within conditions engineers define
A deep-learning model can be accurate and still hard to rely on: its reasoning is hidden, it does not say when a situation falls outside what it learned. Conventional workflows often require another training cycle when model behaviour needs to change. Xpdeep works inside the model itself, across three loops. The customer's system keeps its own rules and authority.
LOOP 01
DESIGN
Read the structure a model has learned and make targeted edits in Native mode, with the rest of the model kept fixed.
LOOP 02
PRODUCTION
Return a prediction with its structural explanation and attributed domain state. The prediction and structural explanation share a forward pass. When useful, provide a constrained minimal feasible intervention with its structural justification, through a separate inference.
LOOP 03
CONTROLLED EVOLUTION
Diagnose operational feedback, make a targeted correction, validate it and redeploy between operating periods. Not continuous learning.
02 THE TEAM AND ITS ORIGIN
From Grenoble research to industrial models
Xpdeep is a research spin-out from the Laboratoire d'Informatique de Grenoble, CNRS and Université Grenoble Alpes. More than two decades of research on temporal data led to the model architecture the team now develops with industrial programmes.
AHLAME DOUZAL · CO-FOUNDER & CSO
Associate professor at Université Grenoble Alpes for 25 years. 60+ publications, 10 PhD theses supervised.
STANISLAS CHESNAIS · CO-FOUNDER & CEO
Built Netsize from zero to €100M net revenue across 23 countries before its sale to Gemalto.
03 THE WORK
Where the work happens
The technology spans these areas. What each role covers is set out in its own listing below.
01
MODEL RESEARCH
Deep-learning architectures whose learned structure can be read and edited.
02
TEMPORAL DATA AND SIGNALS
Multivariate time series and signals, where Xpdeep's edge is deepest.
03
MODEL ENGINEERING
PyTorch-compatible models that train, run and deploy within real constraints.
04
VISUALIZATION
Making learned structure, domain state and interventions legible to engineers.
05
INDUSTRIAL INTEGRATION
Fitting model outputs into the customer's own systems, rules and validation chain.
04 OPEN POSITIONS
Open positions
Roles, locations and contract details come directly from our recruitment platform, in the language each listing is written in. To apply, open a listing and use its application button.