Postdoc in Energy-Efficient AI Math & Hardware

ecolecentraledelyon; Écully; Sur place; EUR 24 000 - 36 000; Plein temps

Avantages offerts par ce poste

Double supervision

Conference travel support

Housing assistance for internationalRe

Résumé du poste

École Centrale de Lyon invites a PhD candidate to work on energy‑efficient diffusion models and score‑based inference. You will quantify score errors, link device physics to theory, and extend guarantees for mixed error patterns.

The project culminates in a co‑design with INL and a proof‑of‑concept on PCM/photonic accelerators, aiming for measurable energy gains. The role demands a strong mathematical background, familiarity with PyTorch or JAX, and excellent English; French is not required.

Qualifications

• A PhD in applied mathematics, statistics, machine learning, electronic engineering, computer science or a closely related field, obtained (or with the defence scheduled) before the start of the contract.
• Strong foundation in probability and stochastic analysis, numerical analysis, optimisation, or machine learning: SDEs, Girsanov’s theorem, time reversal of diffusions.
• Familiarity with convergence theory of sampling algorithms or related analysis through Langevin/MCMC, optimal transport or concentration of measure.
• Solid scientific computing (PyTorch or JAX), and the habit of writing repeatable code.
• Working English; French is not required.

Responsabilités

• Make the score error measurable starting from Gaussian targets with closed‑form error decomposition.
• Map device physics onto the theory and propagate error models into a score‑error budget.
• Extend guarantees to mixed write-time freeze and per-call redraw errors and identify control targets.
• Co‑design with INL and deliver a proof of concept, exploring energy‑accuracy tradeoffs with device‑level models.

Connaissances

Probability & stochastic analysis

Numerical analysis

Optimization

Machine learning

PyTorch/JAX

Formation

PhD in applied mathematics, statistics, machine learning, electronic engineering, computer science or closely related field

Outils

PyTorch/JAX

Description du poste

École Centrale de Lyon invites a PhD candidate to work on energy‑efficient diffusion models and score‑based inference. You will quantify score errors, link device physics to theory, and extend guarantees for mixed error patterns.

The project culminates in a co‑design with INL and a proof‑of‑concept on PCM/photonic accelerators, aiming for measurable energy gains. The role demands a strong mathematical background, familiarity with PyTorch or JAX, and excellent English; French is not required.