Staff Software Engineer, Inference Routing
Коротко о вакансии
PolarGrid, канадская компания на ранней стадии, ищет Staff Software Engineer для разработки и эксплуатации уровня маршрутизации запросов к GPU. Вы будете проектировать и писать код роутера, определять техническое направление и работать в тесной связке с командой из пяти инженеров. Ключевое — глубокое понимание сетей, прокси, балансировщиков нагрузки и распределённых систем, а также опыт написания production-кода на Go или Rust. Опыт работы с GPU или ML не требуется, важны системные знания и желание строить новое с нуля.
- Формат работы
- удалёнка
- Зарплата
- 220 000 — 265 000 CAD — по данным Himalayas
- Ключевые требования
- Глубокие знания сетей, прокси и балансировщиков нагрузки
- Сильные основы распределённых систем: согласованность, обработка отказов, retries, backpressure
- Production-код на системном языке: Go или Rust
- Измерение хвостовой задержки (tail latency) и поиск её источников
- Опыт установки технического направления для других инженеров
- Кому подойдёт
- Вакансия подойдёт опытному инженеру, который хочет с нуля построить критически важный компонент инфраструктуры для AI-инференса.
Пересказ openqareer по тексту работодателя. Ниже — оригинал объявления.
About PolarGrid
PolarGrid is an early-stage Canadian company that runs managed AI inference. Customers bring their own models, and we run them as fast, reliable production endpoints on GPU servers across many sites. We are a small team of about five engineers.
The role
You will own the routing layer: the service that decides which GPU serves each request. You will build it, run it in production and set its technical direction.
This is a hands-on staff role. Most of your time goes to designing and writing the router and the placement logic behind it. You will work closely with the engineers who run our network and our GPU sites, and you will help hire the engineers who join you.
The work
Every request has to land on a GPU that is healthy, has room and is close enough to be fast. The hard part is that capacity and health change constantly across many sites, and the router has to make good decisions quickly with information that is always slightly out of date. It also has to fail safely: when a site or server goes down, traffic moves without customers noticing.
What you bring
-
Deep knowledge of how networks, proxies and load balancers work underneath, from building or running production systems.
-
Strong distributed systems fundamentals: consistency, failure handling, retries and backpressure.
-
You write production code in a systems language such as Go or Rust.
-
You measure latency at the tail, not just the average, and know how to find where it comes from.
-
You have been on call for what you built.
-
You have set technical direction that other engineers built on.
-
You write clearly.
-
You use AI tools well and check their work.
You do not need GPU or ML experience. What matters is that you understand the systems underneath deeply and want to build a new one from the start.
Originally posted on Himalayas
openqareer не кадровое агентство: вакансия собрана с открытого источника, отклик подаётся на сайте работодателя.