AI Engineer, Large-Scale Pre-Training

On-site · Munich (Germany) · Mid · Full-time

Want offers like this one?

Create your free PyHires profile: you will see how well you match each offer and get a weekly email with the ones that fit you.

Create my profile

Overview

Work mode
On-site
Location
Munich (Germany)
Engagement
Full-time
Seniority
Mid
Languages
English
Open to candidates from
EU countries only
Company
Vesoma
Posted
Nov 19, 2026

About the role

Vesoma is building versatile, trustworthy embodied AI for physical labour, with robots engineered to automotive standards. As a key member of the early AI team you will work with the AI leadership and researchers to define the infrastructure used to train large robot foundation models on petabyte-scale datasets across the GPU fleet, balancing stability and research velocity. This is a full-time, permanent position. What you'll do - Develop the software infrastructure to train multi-billion-parameter robot foundation models on large video datasets across multi-node GPU clusters - Build high-throughput data pipelines that stream, decode and preprocess petabyte-scale video, keeping GPUs fully utilised with deterministic, resumable training - Build tools to organise, monitor and optimise large training workloads and sweeps, and remove bottlenecks to maximise goodput What you bring - 3+ years of hands-on experience training multi-billion-parameter models on multi-node GPU clusters, including long-running experiments and in-flight debugging - Deep expertise in distributed training: DDP, FSDP/FSDP2, ZeRO, tensor, pipeline and sequence parallelism, and frameworks such as Megatron-Core, DeepSpeed, TorchTitan or JAX/GSPMD - Training efficiency: mixed precision (BF16/FP8), activation checkpointing and related techniques - Core stack: Python, PyTorch and/or JAX, CUDA, NCCL, and cluster orchestration with Slurm or Kubernetes What they offer - A key role in an early AI team, working with leadership that has built unicorn-scale companies - The chance to help build a European champion in physical AI

Requirements

Based on:Python

Must-have

  • JAX
  • Kubernetes
  • PyTorchEssential

About Vesoma

Vesoma is building humanoid robots for physical labour, engineered to automotive standards of safety, scale and durability, with the goal of becoming a European leader in physical AI. It is well capitalised and has two main locations, in Munich and Limassol.

Stack

  • Python
  • JAX
  • Kubernetes
  • PyTorch

Company