Назад
Company hidden
7 дней назад

Agentic Software Engineer (AI)

Формат работы
onsite
Тип работы
fulltime
Грейд
senior
Английский
b2
Страна
Serbia
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
Agentic Software Engineer (AI): Building production AI agents, multi-agent workflows, harnesses, and quality gates that transform approved requirements into verified software changes with an accent on traceability, security, and responsible AI governance. Focus on designing durable agent loops, engineering MCP-based context and isolated sandboxes, and measuring agent reliability, cost, auditability, and delivered client value.

Location: Belgrade, Serbia

Company

hirify.global is a global transformation partner specializing in technology strategy, business innovation, digital solutions, applications, and device and systems engineering, including regulated health and finance environments.

What you will do

  • Design and build multi-agent workflows with durable state, clear hand-offs, explicit failure handling, stop conditions, retry budgets, and model selection.
  • Engineer agent context and harnesses using tools, domain knowledge, MCP, instruction files, skills, custom agents, hooks, and CI checks.
  • Implement spec-driven development with traceability from requirements to code and tests.
  • Build quality gates combining deterministic checks, evaluations, AI judgment, hidden acceptance tests, evidence, and human decisions.
  • Harden and observe agents through prompt-injection defenses, isolated least-privilege sandboxes, audit trails, cost tracking, and lead-time metrics.
  • Take agent systems into client projects and production, then measure adoption and delivered value after go-live.

Requirements

  • At least seven years of professional software engineering experience and a track record of owning code quality in production systems.
  • Hands-on experience building and using a working agentic system, demonstrated during the selection process as a repository or live demo.
  • Experience with multi-agent orchestration, context engineering, Model Context Protocol (MCP), agent evaluations, prompt-injection guardrails, and agent-run tracing.
  • Experience with agent frameworks such as Claude Agent SDK, Azure AI Foundry, or LangGraph.
  • Experience with AI coding-agent harnesses such as Claude Code, GitHub Copilot, or OpenAI Codex, including instruction files, skills, custom agents, and hooks.
  • Experience building agent loops with independent evaluators, stop conditions, and cost limits, and explaining loop failures.
  • Strong programming skills in Python or TypeScript; clean, tested, well-structured production code is required.
  • Working knowledge of responsible AI and AI governance, including human oversight, logging, and documentation obligations such as those under the EU AI Act.
  • Solid CI/CD and Git practice; confident English communication and effective collaboration with colleagues, client teams, and non-engineers in an international environment.

Nice to have

  • Java or .NET experience.
  • Experience with GitLab, containers, or Kubernetes.
  • Delivery experience in regulated industries such as finance or health, where traceability is mandatory.

Culture & Benefits

  • Full-time role in a global, multidisciplinary engineering and innovation environment.
  • Work directly with client teams and remain accountable for whether delivered capabilities create the expected value.
  • Projects span complex and regulated domains, including health and finance.
  • Applications are welcomed from people of all backgrounds, regardless of protected personal characteristics.

Hiring process

  • Demonstrate a working agentic system during the selection process as a repository or live demo.

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 →