Apertus Engineer: Deployment

Apertus Engineer:
Deployment

100%, Zurich, fixed-term

We are seeking a skilled engineer to own the technical release path of Apertus models. The ideal candidate will integrate Apertus into the open‑source inference ecosystem, produce quantised variants, and make sure each release works out of the box for the community. This role requires strong Python and software engineering skills, experience with LLM inference stacks, and a track record of open‑source contributions.

We train open foundation models with hundreds of billions of parameters on thousands of GPUs on one of the largest AI‑ready supercomputers in Europe. The team counts more than a dozen full‑time engineers working alongside leading researchers from EPFL and ETH Zürich, has released the Apertus 1 and Apertus 1.5 models, and works with over thirty academic collaborators to deliver fully open (open source), responsibly trained, multilingual, multimodal AI models for research and industry.

Apertus is trained and developed on Alps, the Swiss National Supercomputing Centre's (CSCS) supercomputing infrastructure. This role sits at the interface between the training team and the open‑source community: the social side of community engagement is owned by our community manager, while this role owns the technical release path.

The engineer will make Apertus releases work out of the box across the open‑source LLM ecosystem, from server‑grade inference to personal deployment.

• Manage the technical release path of trained Apertus models: checkpoint conversion and preparation of release artefacts (weights, configurations, tokenisers, model cards) together with the training team
• Implement and upstream support for Apertus model architectures in community libraries such as Hugging Face Transformers, vLLM, SGLang, and llama.cpp, and shepherd these contributions through review so community support is a given
• Verify day‑0 compatibility of new releases with the major inference engines and model formats
• Coordinate release timing and technical materials with the community manager

• Produce quantised variants of released models (e.g. FP8, INT4/AWQ/GPTQ, GGUF) suitable for server and personal deployment
• Validate quantised variants against evaluation benchmarks to ensure quality is preserved

• Provide example scripts and reference configurations showing how to serve and use Apertus models with vLLM, SGLang, and Transformers
• Support personal and local deployment ecosystems such as LM Studio, Ollama, and llama.cpp
• Maintain deployment documentation and troubleshooting guides for the community

• MSc or PhD in Computer Science, Data Science, Artificial Intelligence, Machine Learning, or a related field
• Exceptional BSc candidates with strong engineering experience will also be considered
• Strong Python and software engineering skills, including experience with open‑source contribution workflows (pull requests, code review, CI)
• Experience with LLM inference stacks such as Hugging Face Transformers, vLLM, or SGLang
• Strong collaboration and communication skills and ability to work across research, engineering, and community‑facing teams
• Prior hands‑on experience in the core domains of this role is required
• This can be project or study based experience; formal work experience is preferred
• A high degree of flexibility: priorities, tools, and day‑to‑day tasks shift with training schedules, releases, and a fast‑moving field
• A track record of merged contributions to ML or inference libraries (e.g. Transformers, vLLM, SGLang, llama.cpp)

• Experience converting models between formats and frameworks (e.g. Megatron‑LM checkpoints, safe tensors, GGUF)
• Familiarity with personal and local deployment tools such as LM Studio, Ollama, or llama.cpp
• Experience writing developer‑facing documentation and example code

• Published research in the domains relevant to this role, or familiarity with recently published research on these topics
• Experience quantising models without performance degradation (FP8, INT4, AWQ, GPTQ) and evaluating quantised models
• Experience…