Remote job
LLM Engineer — Post-Training (Remote - UK)
Job details
About this role
Role overview Improve the behavior of open-weight language models used in a consumer-facing AI product. You’ll work across training data, post-training, evaluation, and production, turning observed failures into experiments and improvements to model quality.
Responsibilities - Fine-tune open-weight models with supervised fine-tuning, LoRA/QLoRA, and methods such as DPO or GRPO. - Investigate model failures and choose whether to address them through data, training, prompts, or classifiers. - Build training datasets and evaluations that guide subsequent model iterations. - Improve personalization, consistency, tone, multilingual behavior, and group interactions. - Deploy model improvements and use real-world behavior to inform follow-up work. - Make training experiments faster, comparable, and reproducible.
Requirements - Hands-on experience post-training open-weight LLMs for a user-facing product and deploying them to production. - Experience in a business where AI is central to the product. - Strong Python, PyTorch, and Hugging Face skills, including GPU training and debugging. - Experience creating data and evaluations and using evidence to make decisions. - Ability to carry work from experimentation through production. - Comfort evaluating uncensored models and explicit or harmful content as part of model and moderation development.
Nice to have - Conversational, roleplay, or personalized consumer AI experience; multilingual or multimodal model work; moderation, safety, or adversarial testing; open-source LLM contributions.
Benefits and work setup Fully remote, with a preference for B2B contracting and flexibility on contract type. Includes 20 working days of annual paid leave, a yearly team gathering, health and wellness support, coworking and learning budgets, equipment funding, and access to AI tools.