Principal Engineer, Duo Agent Platform (Python, TypeScript)
GitLab is the intelligent orchestration platform for DevSecOps. GitLab enables organizations to increase developer productivity, improve operational efficiency, reduce security and compliance risk, and accelerate digital transformation. Mo…
↑ Вакансия с автоподнятиемGitLab is the intelligent orchestration platform for DevSecOps. GitLab enables organizations to increase developer productivity, improve operational efficiency, reduce security and compliance risk, and accelerate digital transformation. More than 50 million registered users and more than 50% of the Fortune 100* trust GitLab to ship better, more secure software faster.
The same principles built into our products are reflected in how our team works: we embrace AI as a core productivity multiplier, with all team members expected to incorporate AI into their daily workflows to drive efficiency, innovation, and impact. GitLab is where careers accelerate, innovation flourishes, and every voice is valued. Our high-performance culture is driven by our values and continuous knowledge exchange, enabling our team members to reach their full potential while collaborating with industry leaders to solve complex problems. Co-create the future with us as we build technology that transforms how the world develops software.
*Fortune 500® is a registered trademark of Fortune Media IP Limited, used under license. Claim based on GitLab data. Fortune 100 refers to the top 20% ranked companies in the 2025 Fortune 500 list, published in June 2025. Fortune and Fortune Media IP Limited are not affiliated with, and do not endorse products or services of GitLab.
About the role
We are looking for a Principal Engineer to help build GitLab Duo Agent Platform, the agentic orchestration layer that runs AI agents and flows across the entire software lifecycle.
Principal Engineers at GitLab are recognized experts across multiple technology domains and represent the senior technical leadership within their organization. In this role you will set technical direction for a significant part of the Agent Platform, working across several teams to turn agentic capabilities into something engineering organizations can actually depend on: predictable, observable, governable, and fast enough to sit in the critical path of real development work.
The Agent Platform is not a single service. It spans agent execution and orchestration, flow definition and event-driven triggering, the AI Catalog through which customers discover and manage agents, model routing and evaluation, and the governance surface that decides which agents may act where and under whose identity. It runs on GitLab.com, GitLab Dedicated, and Self-Managed installations, including deployments that use self-hosted models to meet compliance requirements. You will work on the parts of that system where the hard architectural questions live, and you will do it hands-on.
This role reports into the VP of the AI Engineering organization and works closely with Engineering Managers, Staff Engineers, Product Management, Infrastructure and Security.
What you'll do
- Define the technical direction for a significant area of Duo Agent Platform across several teams, and partner with Engineering Managers and Product to turn it into a roadmap with a six-month horizon.
- Build the orchestration layer that makes agents and flows reliable: session lifecycle, tool invocation, context assembly, concurrency, retries, timeouts, and failure modes that degrade gracefully instead of stalling a developer's workflow.
- Hands on work on driving topics throughout the whole organisation from POC level, architecture to full implementation.
- Participate in Incident Management on-call rotation, to help meet availability goals for GitLab.com.
- Strong product builder and ownership mentality. Understanding of product strategy, technical strategy, delivering customer value and iteration towards a successful product
- Write clear, opinionated design documents, architecture narratives, and decision records that let other teams make aligned decisions. Ensure the architecture holds across multi-tenant GitLab.com, Dedicated, and Self-Managed deployments, including self-hosted model configurations, and respects data governance boundaries in each.
- Use the platform on our own work, find where it falls short of what we claim, and feed that back through improvements into the product.
- Mentor Staff and Senior Engineers, and raise the bar for experimentation rigor and cross-team collaboration.
- Work across the whole platform on different projects, features and teams. This reaches across:
- Design the extensibility surface so customers can build custom agents and flows, connect their own tools and internal systems, and bring external agents such as Claude Code and Codex into GitLab workflows without weakening the platform's guarantees.
- Drive down the cost and latency of agentic requests, and work with Product on how model selection, routing, and credit consumption behave as usage grows.
- Define the evaluation practice for the platform: offline benchmarks and online experiments that measure task completion, correctness, tool-call accuracy, latency, and cost per task, and that catch regressions before customers do.
- Build the governance and safety surface with Security and Compliance: policy-driven controls over which agents and models may run where, permission and identity semantics that align to existing roles and group structure, human-in-the-loop checkpoints, traceability of every flow session, and audit that stands up to enterprise review.
- Complete Interview Training and participate in hiring for the department as a technical interviewer or panel member.
What you'll bring
- 8+ years of software engineering experience, including 3+ years at Staff, Principal level or equivalent senior technical leadership.
- Hands-on production experience building on large language models: agent loops, tool calling, context management, retrieval, prompt and model evaluation. Experience shipping this to real users, not only prototyping it.
- Strong background in distributed systems, including service decomposition, fault tolerance, observability, and operational resilience under multi-tenant load.
- A track record of defining evaluation frameworks and using benchmark and experiment results to make architecture decisions, rather than shipping on intuition.
- Experience designing human-in-the-loop controls, safety guardrails, and responsible AI practices for systems that act on a customer's behalf.
- Ability to drive alignment across Engineering, Product, Infrastructure, and Security on problems that start out ambiguous.
- Experience mentoring senior engineers and shaping technical direction across multiple teams without direct authority.
- Comfort working in one or more of Ruby, Go, Python, or TypeScript, and willingness to work across the stack the platform spans.
- Excellent written and asynchronous communication, and the ability to work effectively in a fully remote, globally distributed organization.
About the team
The AI engineering organisation is building the product and infrastructure that make GitLab an effective environment for running agents across the software development lifecycle. We’re distributed across the world and partner with teams throughout GitLab to help the GitLab Duo Agent Platform become the backbone of GitLab's AI features. You will report directly to the VP of Engineering for AI and especially work together with all the teams, the engineering management and other principal engineers in the AI department. This is an opportunity to lead a consequential group as AI changes how software is built.
The base salary range for this role’s listed level is currently for residents of the United States only. This range is intended to reflect the role's base salary rate in locations throughout the US. Grade level and salary ranges are determined through interviews and a review of education, experience, knowledge, skills, abilities of the applicant, equity with other team members, alignment with market data, and geographic location. The base salary range does not include any bonuses, equity, or benefits. See more information on our benefits and equity. Sales roles are also eligible for incentive pay targeted at up to 100% of the offered base salary.
How GitLab Supports Full-Time Employees
- Benefits to support your health, finances, and well-being
- Flexible Paid Time Off
- Team Member Resource Groups
- Equity Compensation & Employee Stock Purchase Plan
- Growth and Development Fund
- Parental Leave
Please note that we welcome interest from candidates with varying levels of experience; many successful candidates do not meet every single requirement. Additionally, studies have shown that people from underrepresented groups are less likely to apply to a job unless they meet every single qualification. If you're excited about this role, please apply and allow our recruiters to assess your application.
Country Hiring Guidelines: GitLab hires new team members in countries around the world. All of our roles are remote, however some roles may carry specific location-based eligibility requirements. Our Talent Acquisition team can help answer any questions about location after starting the recruiting process.
Privacy Policy: Please review our Recruitment Privacy Policy. Your privacy is important to us.
GitLab is proud to be an equal opportunity workplace and is an affirmative action employer. GitLab’s policies and practices relating to recruitment, employment, career development and advancement, promotion, and retirement are based solely on merit, regardless of race, color, religion, ancestry, sex (including pregnancy, lactation, sexual orientation, gender identity, or gender expression), national origin, age, citizenship, marital status, mental or physical disability, genetic information (including family medical history), discharge status from the military, protected veteran status (which includes disabled veterans, recently separated veterans, active duty wartime or campaign badge veterans, and Armed Forces service medal veterans), or any other basis protected by law. GitLab will not tolerate discrimination or harassment based on any of these characteristics. See also GitLab’s EEO Policy and EEO is the Law. If you have a disability or special need that requires accommodation, please let us know during the recruiting process.
Похожие вакансии MLOps инженер
Мы - Управление развития искусственного интеллекта блока "Развитие клиентского опыта B2C". В нашей зоне ответственности ML-решения, результаты работы которых ежедневно влияют на поведение десятков миллионов клиентов.
Развитие клиентского опыта – динамический и активно развивающийся процесс. Совместно с бизнесом мы проводим десятки различных экспериментов ежемесячно, и лучшие из них оформляем в полноценные ML продукты. Совокупные финансовые эффекты от наших продуктов > 1 млрд. руб. в год.
Обязанности
- проектировать и развивать end-to-end пайплайны для ML-решений в рамках программы Лояльности: от поставки данных и обучения до A/B-тестирования и вывода моделей в production
- декомпозировать и рефакторить существующие решения (в том числе полу-ручные MVP) в высоконагруженные, отказоустойчивые микросервисы на Python
- тестировать и оптимизировать PySpark ETL/ELT-пайплайны: профилировать узкие места, настраивать джобы, работать с партиционированием и форматом хранения данных для обеспечения производительности на больших объёмах
- разрабатывать внутренние Python-библиотеки и стандарты разработки для переиспользования кода командами Data Science.
Требования
- опыт от 3 лет в разработке на Python, из них от 2–3 лет в роли MLOps Engineer, Data Engineer или Backend-разработчика в ML-платформах
- уверенная работа с экосистемой Big Data: опыт работы, тестирования и оптимизации PySpark-пайплайнов, Apache Airflow (или аналоги), Kafka, Docker/Kubernetes (K8s), S3/HDFS
- опыт проектирования архитектуры ML-решений: понимание микросервисов, асинхронности, паттернов проектирования
- понимание полного цикла жизни ML-модели: от подготовки данных и feature engineering до мониторинга в проде (MLflow, Prometheus/Grafana или аналоги).
будет плюсом:
- опыт автоматизации CI/CD для полного цикла ML (обучение, валидация, развёртывание моделей) с использованием GitLab CI или Jenkins
- понимание PySpark на уровне администратора/архитектора: тонкая настройка кластеров, управление ресурсами (динамическое распределение Dynamic Allocation), работа с executors, partitions, управление памятью (spark.memory.fraction, spark.memory.storageFraction), тюнинг shuffle
- опыт работы с Feature Store, MLflow, ClickHouse.
Условия
- комфортный современный офис рядом с м. Кутузовская
- офисный формат работы
- ежегодный пересмотр зарплаты, годовая премия
- корпоративный спортзал и зоны отдыха более 400 образовательных программ СберУниверситета для профессионального и карьерного развития
- программа адаптации и помощь руководителя на старте
- расширенный ДМС, льготное страхование для семьи и корпоративная пенсионная программа
- гибкий дисконт по ипотечному кредиту, равный 1/3 ключевой ставки ЦБ
- подписка Прайм с возможностью совместного использования на трёх близких
- вознаграждение за рекомендацию друзей в команду Сбера.
МТС Web Services (MWS) - бигтех-компания, предоставляющая облачные, AI-сервисы и платформенные решения под разные задачи бизнеса: от работы с данными до разработки продуктов и оптимизации процессов. Также мы запустили новую публичную платформу собственной разработки — MWS Cloud Platform.
xOps Copilots – это технологическое решение, представляющее собой единую платформу для разработки, внедрения и масштабирования интеллектуальных цифровых агентов и копилотов. Эти агенты — автономные программные сущности на базе ИИ, способные выполнять задачи, принимать решения, обучаться и адаптироваться в реальном времени для автоматизации внутренних бизнес-функций.
Задачи
проектировать архитектуру и программировать пайплайны для бизнес-копайлотов;
самостоятельно разрабатывать решения от прототипа до вывода в пром: реализация логики, обработка ошибок, масштабирование и обеспечение отказоустойчивости;
разрабатывать и дорабатывать LLM/RAG-копайлоты: реализация сложной агентской логики, оркестрация вызовов, управление контекстом;
работать с неструктурированными источниками (PDF, Excel, сканы договоров, локальные нормативные акты);
реализовывать RAG-системы: настройка пайплайнов индексации документов, работа с эмбеддингами, управление векторными базами данных;
при необходимости - дообучение или адаптация моделей под специфику домена.
Требования
опыт разработки на Python, работа с LLM API (OpenAI, Claude или аналогичными);
практический опыт построения RAG-систем: индексация документов, эмбеддинги, векторные БД;
опыт интеграции с корпоративными системами через API (SAP, OEBS/1С, коннекторы Confluence/Jira и т.п.);
понимание основ аналитики эффекта: как считать ROI/эффект внедрения, работа с метриками ;
умение самостоятельно доводить проект от MVP до прод-эксплуатации.
Условия
Собственную платформу MTS Ocean для получения ИТ-ресурсов, а это значит, что деплой, мониторинг, observability - не будут для вас проблемой, вы сможете сосредоточиться на фичах;
Профессиональные гильдии инженеров по направлениям, чтобы поддерживать друг друга и обмениваться опытом;
Внутреннюю площадку TechTalks для обмена опытом, дискуссий, развития навыков самопрезентации;
Участие во внешних IT конференциях. Наши эксперты выступают на ведущих профильных конференциях, таких как HighLoad++, DevOps Conf, Joker, AHA!, Матемаркетинг и другие. А еще мы организуем собственную масштабную конференцию True Tech Day и регулярно проводим митапы для разработчиков и инженеров.
Полезные курсы и вебинары в корпоративном университете и электронные библиотеки;
Медицинскую страховку с 1 месяца со 100% покрытием расходов, включая стоматологию, страхование жизни и здоровья в поездках за рубеж. А еще можно застраховать родственников с корпоративной скидкой;
Единую подписку МТС Premium — KION light в онлайн-кинотеатре KION, сервис МТС Music, 30 дней бесплатного пользования подпиской OZON Premium;
Скидки и предложения от партнеров на фитнес, занятия английским и прочее;
Отпуск 28 календарных дней + 3 дополнительных дня за ненормированный рабочий день;
Гибридный и удаленный формат.
HireSeeker собирает вакансии со всех площадок и присылает только релевантные. Бесплатно.