PhD in Frugal, Resource-Aware Distributed AI Training

1000scholars; Valbonne; On-site; EUR 26,000 - 34,000; Full time

Job summary

1000scholars invites applications for a PhD thesis hosted at the Inria Centre, Sophia Antipolis, focusing on distributed training algorithms for volatile resources within the Hivenet framework.

The project involves collaboration with Inria NEO/ARGO teams, exploring resource availability modeling, energy-aware training, and theoretical guarantees to advance scalable, efficient federated-like learning on dynamic infrastructures.

Qualifications

• Solid mathematical background suitable for theoretical and empirical study.
• Strong interest in distributed systems and resource-aware ML.
• Good programming skills in Python; experience with ML frameworks preferred.
• Fluency in English and ability to communicate research clearly.

Responsibilities

• Research on distributed training algorithms for volatile resources.
• Contribute to modeling resource availability and scheduling strategies.
• Develop and validate algorithms with simulations and experiments in ML systems.

Skills

Mathematical background

Optimization

Probability

Stochastic processes

Statistical machine learning

Distributed systems

Python programming

PyTorch

TensorFlow

JAX

Energy-aware computing

Fluency in English

Tools

PyTorch

TensorFlow

JAX

Job description

1000scholars invites applications for a PhD thesis hosted at the Inria Centre, Sophia Antipolis, focusing on distributed training algorithms for volatile resources within the Hivenet framework.

The project involves collaboration with Inria NEO/ARGO teams, exploring resource availability modeling, energy-aware training, and theoretical guarantees to advance scalable, efficient federated-like learning on dynamic infrastructures.