LLM Engineer: Production-Ready Domain Models

US Health Partners, LLC; Amsterdam; Hybrid; EUR 90,000 - 140,000; Full time

Job summary

Prosus is seeking an experienced Machine Learning Engineer to build domain-specific language models. You will own the full training pipeline—from continued pre-training to fine-tuning on proprietary data—and scale training across multi-node GPU clusters.

You will implement post-training methods like RLHF/DPO, craft robust evaluation frameworks, and ensure production-grade delivery with strong software engineering practices. Hybrid work model available in Amsterdam.

Qualifications

• 3+ years of ML engineering experience.
• Proven experience training and deploying language models to production.
• Experience preparing large-scale training datasets: filtering, quality assessment, deduplication, and data mixture design.
• Hands-on experience with distributed training frameworks (DeepSpeed, FSDP, Megatron-LM, Axolotl).
• Strong understanding of training dynamics at scale: debugging loss instabilities and LR schedules.
• Expert Python and PyTorch with production experience using Transformers, DeepSpeed, Accelerate.

Responsibilities

• Analyze model performance and data to identify improvements.
• Design, run, and evaluate experiments to improve model quality and efficiency.
• Train large language models through pre-training and fine-tuning on proprietary data.
• Build post-training pipelines using RLHF, DPO, GRPO and similar methods.
• Develop distributed training infrastructure across multi-node GPU clusters.
• Prepare and curate 100B+ token-scale training corpora.
• Generate synthetic data for instruction tuning and enhancements.
• Debug stability, optimize throughput, monitor performance in long-running jobs.
• Implement techniques from recent research adapted to our use cases.
• Create robust evaluation frameworks to guide training decisions.
• Write production-grade code and contribute to engineering standards.

Skills

ML engineering

Language model training

Distributed training

Python

PyTorch

Transformers

CI/CD

Tools

DeepSpeed

FSDP

Megatron-LM

Axolotl

Job description

Prosus is seeking an experienced Machine Learning Engineer to build domain-specific language models. You will own the full training pipeline—from continued pre-training to fine-tuning on proprietary data—and scale training across multi-node GPU clusters.

You will implement post-training methods like RLHF/DPO, craft robust evaluation frameworks, and ensure production-grade delivery with strong software engineering practices. Hybrid work model available in Amsterdam.