Назад
Company hidden
7 дней назад

Post-training Specialist (LLM/NLP)

Формат работы
onsite
Тип работы
fulltime
Английский
c1
Страна
Spain
Вакансия из списка Hirify.GlobalВакансия из Hirify Global, списка международных tech-компаний
Для мэтча и отклика нужен Plus

Мэтч & Сопровод

Для мэтча с этой вакансией нужен Plus

Описание вакансии

Текст:
/
TL;DR
Post-training Specialist (LLM/NLP): Designing and implementing post-training techniques, datasets, and evaluation workflows for language models serving Spanish and other Iberian languages with an accent on instruction tuning, reinforcement learning, and linguistic quality. Focus on scaling HPC training workflows, identifying regional-language performance gaps, and integrating linguistic specificities into general-purpose models.

Location: Barcelona, Spain. The position will be located at BSC within the Directors Department.

Salary: Competitive salary commensurate with qualifications and experience and according to the cost of living in Barcelona.

Company

hirify.global is Spain’s leading supercomputing center, combining HPC service provision with research and development in computer and computational science, with more than 1,000 staff from 60 countries.

What you will do

  • Execute and monitor post-training experiments, including instruction tuning and reinforcement learning techniques such as DPO, GRPO, OPD and variants, for Iberian languages.
  • Implement and scale model-training workflows with Slime, TRL and NeMo-RL.
  • Develop and curate datasets in Spanish, Catalan, Basque and Galician.
  • Evaluate model performance across regional linguistic nuances, including generation and reasoning quality.
  • Collaborate with senior engineers to integrate linguistic specificities into general-purpose models in an HPC environment.

Requirements

  • Bachelor’s degree in Computer Science, Mathematics, Statistics, Engineering or a related field, plus a Master’s degree in Artificial Intelligence, Machine Learning, Computer Science, Data Science, Applied Mathematics or a closely related field.
  • Solid experience in fine-tuning and post-training of Large Language Models.
  • Proficiency in Python and deep learning frameworks such as PyTorch.
  • Hands-on experience with modern post-training frameworks such as Slime, Verl or NeMo-RL.
  • Familiarity with NLP tasks and evaluation metrics.
  • Excellent written and spoken English.
  • Strong analytical, collaborative and multidisciplinary research skills, with an innovative, proactive and detail-oriented approach to open science, data quality and model evaluation.

Nice to have

  • Experience with advanced reinforcement learning methods such as GRPO, OPD and their variants.
  • Knowledge of linguistic features and datasets for Spanish and Spain’s co-official languages.
  • Fluency in Spanish.
  • Knowledge of Catalan, Basque or Galician.

Culture & Benefits

  • Full-time open-ended contract, 35 hours per week, linked to the technical and scientific activities and project budget.
  • Flexible working hours and a highly stimulating environment with state-of-the-art infrastructure.
  • Extensive training plan, restaurant tickets and private health insurance.
  • Support with relocation procedures.
  • 22 vacation days, 6 personal days, and 24 and 31 December holidays under the collective agreement.
  • Starting date: 1 December 2026.

Hiring process

  • Submit the application through the BSC website with a full CV in English including contact details.
  • Include an English cover or motivation letter stating the specific area and topics of interest, plus two references; applications without this document will not be considered.
  • Selection comprises curriculum analysis worth 40 points and an interview phase worth 60 points, with at least 30 interview points required. The process includes technical and administrative interviews, a personality questionnaire and a technical exercise.

Будьте осторожны

Если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого. Это мошенники. Обязательно жмите «Пожаловаться» или пишите в поддержку. Подробнее в гайде →