Чётко очерченная роль в AI-стартапе на переднем крае рынка даёт возможность работать с инфраструктурой обучения крупных моделей, но офисный формат и стартап-среда подойдут не всем.
Выше рынка
Передовой AI
Современный стек
Понятная роль
Только офис
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Software Engineer, Training Infrastructure (AI): Building and operating distributed infrastructure for frontier model training and Tinker fine-tuning across large GPU clusters with an accent on scheduling, fault tolerance, throughput, and cost efficiency. Focus on optimizing large-scale model architectures, supporting LoRA and reinforcement learning workloads, and designing recovery and observability systems that withstand hardware and network failures.
Location: Onsite in San Francisco, California or New York, New York, United States
Annual salary: $300,000–$400,000 USD, depending on background, skills, and experience.
Company
Thinking Machines Lab builds AI systems that extend human will and judgment, including frontier models, the Tinker fine-tuning API, and human-AI interfaces.
What you will do
Design, build, and operate distributed training infrastructure for Tinker and internal frontier model training, including scheduling, orchestration, checkpointing, and recovery.
Improve training throughput and cost efficiency through profiling, parallelism strategies, memory optimization, and kernel- and communication-level tuning.
Bring up and optimize support for open-weights and in-house model architectures, including mixture-of-experts models.
Build reliability through automated failure detection, job recovery, multi-tenant isolation, and observability for long-running jobs.
Support multi-tenant LoRA fine-tuning and tightly coupled sampling and training for reinforcement learning workloads.
Collaborate with API, frameworks, and product teams to keep infrastructure and the user-facing SDK consistent and reliable.
Requirements
4+ years of software engineering experience, including building and operating production distributed systems.
Strong Python programming skills plus experience with a systems language such as C++, Rust, or Go.
Hands-on experience with large-scale model training or inference infrastructure using frameworks such as PyTorch.
Demonstrated ability to debug complex performance and reliability issues in distributed environments.
Nice to have
Experience training or serving large language models at scale, including mixture-of-experts architectures.
Experience with LoRA and other parameter-efficient fine-tuning methods, or supervised and reinforcement learning pipelines for LLMs.
Experience with cluster orchestration and scheduling, such as Kubernetes or Slurm, and fault-tolerant checkpointing.
Experience building observability and profiling tooling.
Contributions to open-source machine learning systems projects such as PyTorch, Megatron, vLLM, SGLang, or similar.
Culture & Benefits
Full-time role in the Core Engineering department.
Visa sponsorship is available, with support through the visa process when appropriate.
Health, dental, and vision benefits.
Unlimited paid time off and paid parental leave.
Relocation support as needed.
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 →