Senior Machine Learning Engineer (LLMs)

Location: Hybrid work in Amsterdam, Netherlands

Company

hirify.global is a major consumer internet company and technology investor building AI solutions that support online commerce and other high-impact applications.

What you will do

• Train domain-specific language models through continued pre-training and full-parameter fine-tuning on proprietary datasets.
• Design experiments to improve model quality, training efficiency, inference performance, and downstream task results.
• Build and optimize distributed training infrastructure across multi-node GPU clusters using DeepSpeed, FSDP, Megatron-LM, and Axolotl.
• Prepare and curate large-scale training data, including filtering, quality assessment, deduplication, data mixture design, and synthetic data generation.
• Develop evaluation frameworks, debug training stability issues, and optimize model serving through quantization, distillation, and throughput improvements.
• Provide technical leadership through architecture decisions, code reviews, mentoring, and engineering standards.

Requirements

• 7+ years of machine learning engineering experience.
• Proven experience training and deploying language models to production, including pre-training, continued pre-training, or fine-tuning.
• Hands-on experience with distributed training across multi-node jobs and strong knowledge of training dynamics and stability at scale.
• Expert Python and PyTorch skills with production experience using Transformers, DeepSpeed, and Accelerate.
• Experience preparing large-scale training datasets and designing filtering, deduplication, quality assessment, and data mixture strategies.
• Technical leadership experience, including mentoring engineers, conducting code reviews, making architecture decisions, and delivering projects with measurable impact.

Nice to have

• Experience with RLHF, DPO, GRPO, or other reinforcement learning approaches for alignment and instruction following.
• Production inference optimization experience with quantization, model compression, distillation, vLLM, or TensorRT-LLM.
• Knowledge of GPU architectures such as A100, H100, and H200, plus memory optimization techniques including gradient checkpointing, mixed precision, and ZeRO.
• Published ML research, open-source contributions, public benchmarks, or models released on Hugging Face.

Culture & Benefits

• Work on strategically important AI projects affecting millions of online shoppers.
• Access to H200 GPU infrastructure, large proprietary datasets, frontier models, and modern ML tooling.
• Significant autonomy to test ideas and drive technical decisions.
• Hybrid work model based around the Amsterdam AI House and its AI community.
• Competitive compensation, a top-spec MacBook Pro, and opportunities for professional growth and learning.