Назад
Company hidden
3 дня назад

Staff or Lead AI Safety Engineer (AI Safety)

Формат работы
onsite
Тип работы
fulltime
Грейд
lead
Английский
b2
Страна
Denmark
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
Staff or Lead AI Safety Engineer (AI Safety): Building safe and secure agentic AI systems through rigorous evaluation, threat and risk modelling, control design, and robust system architectures with an accent on downstream and upstream AI safety. Focus on designing evaluation suites, red-teaming campaigns, LLMOps monitoring and guardrails, and translating ambiguous client requirements into secure technical solutions.

Location: Copenhagen, Kronprinsensgade 7, Denmark

Company

Halfspace, part of Faculty and hirify.global, delivers applied AI and advanced analytics solutions, including AI safety work for frontier and agentic AI systems.

What you will do

  • Build Halfspace’s technical AI Safety capabilities and define best practices for safe and secure agentic AI systems.
  • Design and assess rigorous evaluations, threat and risk models, controls, and robust system architectures.
  • Lead safety-focused client engagements, ensuring quality and rigour in solution designs and assessments.
  • Work with Halfspace leadership, commercial leads, and senior client executives to scope safety engagements and shape the safety offering.
  • Address both Downstream Safety for systems built around foundation models and Upstream Safety for client-trained or fine-tuned models.
  • Mentor junior colleagues and collaborate with engineers and AI Safety experts from Faculty and hirify.global.

Requirements

  • Proven track record of leading AI Safety work or adjacent work, including agentic-system evaluation design, AI red teaming, threat and risk modelling, or control design and implementation.
  • MSc or PhD in computer science, machine learning, mathematics, statistics, physics, engineering, or another STEM field, or equivalent practical experience.
  • Experience designing and implementing evaluation suites and red-teaming campaigns for agentic AI systems using rigorous, quantitative assessment and communication of results.
  • Experience in LLMOps, including online monitoring, guardrails, and controls for agentic AI systems.
  • Ability to build secure cloud-based agentic AI systems using Azure, AWS, or GCP, with a foundation in classic security engineering.
  • Understanding of AI-system threat classes such as prompt injection and tool misuse, with the ability to translate them into technical designs, quantitative evaluations, and metrics.
  • Proficiency in Python, including agentic AI frameworks such as Pydantic AI and LangGraph and evaluation frameworks such as Inspect AI.
  • Ability to turn ambiguous business requirements into defined solution scopes, weigh trade-offs, and explain technical choices, risks, and trade-offs to technical and non-technical audiences.
  • Ownership, quality focus, curiosity, and willingness to share knowledge and help colleagues grow.

Nice to have

  • Consulting experience delivering results quickly in ambiguous situations.
  • Knowledge of relevant regulatory and audit frameworks, including the EU AI Act, ISO 42001, or AIUC-1.
  • Experience in technical interpretability and alignment of foundation models.
  • PhD or published research in AI safety, machine learning, security, or a closely related field.

Culture & Benefits

  • Full-time role at Staff or Lead level, depending on experience and qualifications.
  • Work alongside engineers and Safety experts from Halfspace, Faculty, and hirify.global.
  • Collaborative environment focused on curiosity, boldness, pioneering work, ownership, and high-quality outcomes.
  • Access to work across applied AI, advanced analytics, AI Safety, and frontier model evaluation.

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 →