Lead LLM Engineer - Hybrid Amsterdam, GPU Training

Bluebird; Amsterdam; Hybrid; EUR 120,000 - 180,000; Full time

Benefits offered by this job

Relocation assistance

Visa support

Job summary

Bluebird in Amsterdam is building a world-class AI team focused on language models. You will train LLMs with continued pre-training and full fine-tuning on massive proprietary datasets, and you’ll design and optimize distributed training across multi-node GPU clusters.

You’ll generate synthetic data for instruction tuning and help shape evaluation frameworks that guide product decisions. The role combines hands-on coding with technical leadership, mentoring ML engineers, and making architecture

Qualifications

• 7+ years in ML engineering with production experience in language models or related models.
• Hands-on expertise in multi-node training, solving loss spikes and timeouts.
• Expertise in Python and PyTorch; experience with DeepSpeed and distributed training.
• Experience as a technical lead, mentoring engineers and delivering business-focused results.
• Nice to have: RLHF, vLLM, distillation, NeurIPS/ICML/ACL/EMNLP papers.

Responsibilities

• Train LLMs with continued pre-training and full fine-tuning on large proprietary datasets.
• Build and optimise distributed training on multi-node clusters (H200, DeepSpeed, FSDP).
• Own data at scale: filtering, deduplication, quality and data mixtures for corpora of 100B+ tokens.
• Generate synthetic data for instruction tuning and build evaluation frameworks.
• Lead the technical work: architecture decisions, code reviews and mentoring ML engineers.

Skills

ML engineering

Distributed training

Python

PyTorch

DeepSpeed

Mentoring

Education

Tools

DeepSpeed

NCCL

Megatron-LM

Transformers

Accelerate

Job description

Bluebird in Amsterdam is building a world-class AI team focused on language models. You will train LLMs with continued pre-training and full fine-tuning on massive proprietary datasets, and you’ll design and optimize distributed training across multi-node GPU clusters.

You’ll generate synthetic data for instruction tuning and help shape evaluation frameworks that guide product decisions. The role combines hands-on coding with technical leadership, mentoring ML engineers, and making architecture