Research Engineer: Architecting Agentic LLMs & VLMs

H Company; Paris; Hybrid; EUR 120,000 - 180,000; Full time

Benefits offered by this job

Competitive salary

Hybrid work model

Multicultural team

Job summary

H Company in Paris or London is seeking a senior research engineer to train and improve large language models and vision-language models, with expertise in distributed training and multimodal architectures.

You will implement SFT, RLHF/RLVR, DPO, reward modelling, and work across data pipelines, training backends, and RL algorithms in a collaborative, fast-paced startup.

Hybrid role requires in-office presence roughly three days per week, with occasional travel between offices.

Qualifications

• PhD or MSc in ML, DL, NLP or CV.
• Experience training LLMs/VLMs at scale.
• Strong software engineering and Python/Rust.
• Ability to work across data ingestion to ML/RL design.

Responsibilities

• Develop and train LLMs and VLMs with multimodal architectures.
• Research and implement training methods for instruction following and tool use.
• Design data pipelines and training systems for large-scale distributed training.
• Collaborate with cross-functional teams to integrate models into agentic systems.
• Evaluate model performance and report findings.
• Stay current with advances in LLMs, VLMs, and related fields.

Skills

Python

Rust

Distributed training

LLMs/VLMs

PyTorch/JAX/TF

SFT

RLHF/RLVR

reward modelling

Education

PhD or MSc in ML/NLP/CV

Tools

PyTorch

JAX

TensorFlow

MoE

Distributed Training

Job description

H Company in Paris or London is seeking a senior research engineer to train and improve large language models and vision-language models, with expertise in distributed training and multimodal architectures.

You will implement SFT, RLHF/RLVR, DPO, reward modelling, and work across data pipelines, training backends, and RL algorithms in a collaborative, fast-paced startup.

Hybrid role requires in-office presence roughly three days per week, with occasional travel between offices.