Senior Machine Learning Engineer

remotehiro; Jena; On-site; EUR 250,000 - 450,000; Full time

Benefits offered by this job

Top-of-market compensation

Ground-up function with exec support

On-site collaboration in Jena

Job summary

remotehiro is seeking an expert to design, train, and scale state-of-the-art Large Language Models and RL systems. You will work with 500+ GPU clusters, implementing RLHF and advanced self-play techniques.

The role emphasizes multi-modal architectures, high-performance optimization, and deployment-ready inference in a cutting-edge AI OS project. The position, based in Jena, Germany, requires relocation after onboarding and offers top-of-market compensation with a fast-paced startup culture

Qualifications

• Expertise in building and training large-scale LLMs.
• Experience with 500+ GPU distributed training.
• Strong RL knowledge and self-play methods.
• Relocation to Jena, Germany after onboarding.

Responsibilities

• Design, train, and optimize Large Language Models (LLMs) from scratch.
• Scale distributed training on massive GPU clusters (500+ H100 GPUs).
• Implement advanced reinforcement learning techniques (RLHF, adversarial self-play, real-time control).
• Develop high-performance architectures for multi-modal AI systems.
• Build simulation environments for RL-based AI agents.
• Optimize inference speed and efficiency for real-world deployment.

Skills

LLM design

Distributed training

RL & self-play

Deep learning optimization

PyTorch

TensorFlow

JAX

CUDA

Triton

DeepSpeed

HPC

Tools

CUDA

Triton

DeepSpeed

Job description

You will push the limits of AI by designing, training, and scaling state-of-the-art Large Language Models (LLMs) and advanced reinforcement learning (RL) systems. We are looking for an exceptional expert who has already built and deployed large-scale LLMs and understands every detail of the process—from tokenization to training at scale.

Your work will drive our next-gen AI models, shaping the future of machine intelligence beyond traditional paradigms. You will work with massive compute clusters (500+ H100 GPUs) and cutting-edge reinforcement learning techniques to create highly efficient, scalable, and groundbreaking AI systems.

Your profile

• Deep expertise in LLMs – you've built and trained large-scale models yourself
• Experience with large-scale distributed training on 500+ GPU superclusters
• Deep understanding of reinforcement learning, neural network optimization, and self-play methods
• Expert in PyTorch, TensorFlow, JAX & low-level optimization techniques (CUDA, Triton, DeepSpeed, etc.)
• Familiarity with high-performance computing (HPC, NVLink, InfiniBand, parallel computing)
• Strong publication track record in AI/ML research is a plus
• Relocation to Jena, Germany after initial onboarding

Responsibilities

• Design, train, and optimize Large Language Models (LLMs) from scratch
• Scale distributed training on massive GPU clusters (500+ H100 GPUs)
• Implement advanced reinforcement learning techniques (RLHF, adversarial self-play, real-time control)
• Develop high-performance architectures for multi-modal AI systems
• Build simulation environments for RL-based AI agents
• Optimize inference speed and efficiency for real-world deployment
• Collaborate with top AI researchers to push the boundaries of machine learning innovation

Company Culture and Work Style

We operate in a dynamic startup environment where speed, efficiency, and innovation are key to achieving our goals and growing together. Our development process is based on rapid iterations, allowing us to quickly implement and test ideas to enhance our product and meet user needs.

What we offer:

• Impact: Own and shape security for a category-defining AI OS.
• Compensation: Top-of-market (open) for the right leader.
• Growth: Build the function from the ground up with strong executive support.
• Focus: On-site collaboration in Jena, Germany.

Language requirements:

• English or German