Vision-Language Research Engineer (Post-Train) - London

Satorus Group; Greater London; Hybrid; GBP 70,000 - 120,000; Full time

Benefits offered by this job

Restaurant d'entreprise

Job summary

Satorus Group in London is seeking a VLM Research Engineer to take open-weight vision-language models from data preparation through post-training and production deployment. You will fine-tune with supervised, RL methods, and establish robust evaluation benchmarks to prove improvements before shipping.

You’ll deploy on AWS, monitor model behavior with LangSmith, and collaborate with engineers to deliver reliable image and video analysis capabilities for enterprise investigations.

Qualifications

• Post-training a model with real user data and moving it to production.
• Rigorous evaluation with held-out sets, baselines and leakage checks.
• Strong research instincts and engineering discipline; readable code is essential.
• Clear communication to engineers and non-engineers about results.

Responsibilities

• Own specialist vision-language models from base to ship.
• Run supervised fine-tuning, preference optimisation and RL with verifiable rewards.
• Build held-out benchmarks and baselines to evaluate model performance.
• Deploy and monitor models on AWS, keeping latency and costs in check.

Skills

Post-training

Python

PyTorch

JAX

AWS

LangSmith

Langfuse

Weights & Biases

Arize

SageMaker

Bedrock

S3-based data pipelines

vLLM

Tools

SageMaker

Bedrock

S3-based data pipelines

vLLM

Job description

Satorus Group in London is seeking a VLM Research Engineer to take open-weight vision-language models from data preparation through post-training and production deployment. You will fine-tune with supervised, RL methods, and establish robust evaluation benchmarks to prove improvements before shipping.

You’ll deploy on AWS, monitor model behavior with LangSmith, and collaborate with engineers to deliver reliable image and video analysis capabilities for enterprise investigations.