Member of Technical Staff – Model Training

Sferical AI; Stockholms kommun; On-site; SEK 1,100,000 - 1,600,000; Full time

Job summary

Sferical AI is seeking an experienced Model Training specialist to customize large language models for engineering and technical use cases. The role is based in Stockholm.

You will own end-to-end model training workflows, from data preparation to deployment, and work with NVIDIA Megatron-based frameworks and RL methods to deliver cutting-edge AI capabilities for partners.

Qualifications

• Extensive experience customizing LLMs for out-of-distribution tasks.
• Hands-on with continued pre-training, supervised fine-tuning and RLHF/RLVR.
• Experience with data engineering for model training on cloud.
• Experience building end-to-end pipelines including inference and evals.
• Proficiency with NVIDIA stack and related frameworks.
• Fluency in English; Swedish is a strong plus; Swedish citizenship mandatory.

Responsibilities

• Customizing LLMs for engineering and technical tasks outside training distribution.
• Leading model training through pre-training, fine-tuning and RL methods.
• Overseeing data engineering on a cloud platform for model training.
• Designing end-to-end pipelines from data to inference and evaluation.
• Collaborating with partner teams to ensure success and integration.

Skills

LLM customization

Continued pre-training

Supervised fine-tuning

Reinforcement learning

Data engineering

Cloud pipelines

End-to-end inference

NVIDIA Megatron

English fluency

Swedish language

Swedish citizenship

Tools

Megatron

Nemotron

NVIDIA stack

Job description

Sferical AI was founded by AstraZeneca, Ericsson, Saab, SEB and Wallenberg Investments to give Swedish industry secure, sovereign computing capacity for AI. Our first supercomputer, Sferical One, is built on NVIDIA DGX SuperPOD architecture. We’re looking for an experienced LLM training specialist to customize large language models for our partners’ engineering and technical use cases.

You’ll join a team whose mission is to drive Sferical AI’s success through workload leadership, helping our customers reach their goals for sovereignty and competitiveness with AI. You’ll also help build Sferical AI into a company that embodies the opportunity AI presents, running on a modern technology platform that integrates our own AI capability, both in our technical mission and in our thought leadership within the Wallenberg sphere.

As a Member of Technical Staff, you are:

• a recognized peer in a highly skilled group with collective responsibility for our platform.
• expected to contribute across the platform design cycle, from strategy and architecture to design, validation and support, and across every part of the platform where you have experience, from hardware and installation to compute, storage, orchestration and security.
• expected to take particular ownership of the areas where your expertise is strongest, which for this role means model training and customization.
• invested in our customers’ success, contributing to integration and troubleshooting, and occasionally acting as lead or single point of contact toward a customer or partner.
• encouraged to contribute in all areas together with your peers, without being expected to act autonomously in all of them.
• free to step outside narrow role descriptions such as developer, data scientist or architect, and to grow your skills in line with our customer commitments and the team’s day-to-day work.
• responsible for producing work your peers can build on, with clear design, architectural consistency, good documentation and solid workmanship.
• expected to lead by example, speak up and help keep the team psychologically safe.

What you’ll own as our Model Training specialist

• customizing LLMs for engineering and technical tasks that fall outside their original training distribution.
• training through continued pre-training, supervised fine-tuning and reinforcement learning (RLVR and RLHF).
• the data engineering behind this work, on a cloud platform.
• end-to-end pipelines, from data and training through to inference and evaluation.
• working closely with use cases and initiatives across our partner companies.

What you bring

• extensive experience customizing LLMs for out-of-distribution engineering or technical task sets.
• hands-on experience with continued pre-training, supervised fine-tuning and reinforcement learning (RLVR and RLHF).
• experience with data engineering for model training on a cloud platform.
• experience building end-to-end pipelines, including inference and evals.
• experience with the NVIDIA stack, such as Megatron-based frameworks and Nemotron models.
• fluency in English; Swedish is a strong plus. Swedish citizenship is mandatory.

This role is based at our Stockholm office, five days a week. We build best together, and we believe working side by side drives better results.