[Skip to main content](#main-content)

[OpenNash Jobs](https://jobs.opennash.com/)
[Hiring report](/reports/hiring/) · [All Xiaomi jobs](https://jobs.opennash.com/companies/xiaomi/) · [Search jobs](/search/)

Xiaomi · 北京

# AI基建-大模型训练框架开发工程师

北京 

Listed on Xiaomi ’s official careers site. Posted July 30, 2026. Check that it is still open and apply there.
[Apply on Xiaomi ’s careers site →](https://xiaomi.jobs.f.mioffice.cn/index/position/7595847797652785446/detail)

[OpenNash Jobs](https://jobs.opennash.com/) / [Xiaomi ](https://jobs.opennash.com/companies/xiaomi/) / AI基建-大模型训练框架开发工程师

## Job description

- 基于 PyTorch、DeepSpeed、Megatron-LM 等技术，参与大模型训练框架的设计、开发与优化，提升模型训练的效率、稳定性与扩展性，支撑大规模模型的高效训练和部署。 - 解决万卡集群中大模型训练中的关键技术难题，包括分布式训练通信、内存显存优化、数据加载与预处理加速等，保障训练高效稳定，降低资源消耗。 - 开展大模型框架性能评估和调优工作，构建并完善性能监控体系，通过实时监测训练指标，定位性能瓶颈，提出优化方案，确保在不同硬件上达到最优性能。

## Qualifications

- 1–5 年高性能计算 / 分布式训练 / 深度学习系统研发经验；
- 熟练阅读并修改 PyTorch/DeepSpeed/Megatron-LM 核心源码，有线上调优案例；
- 熟练使用 nsight/NCCL profiler 分析任务训练性能瓶颈；
- 熟悉 InfiniBand/RoCEv2 网络拓扑，能独立调优 DP/TP/PP/EP 切分策略，解决大规模下通信-计算重叠问题；
- 具备 Python/C++ 混合开发能力，代码洁癖＋CI/CD 意识。

## At a glance
- Posted by the employer: Jul 30, 2026
- First seen by OpenNash: Aug 1, 2026
- Last checked: Oct 19, 2026
- Listed for: 81 days
- Location: 北京[Apply on Xiaomi ’s careers site →](https://xiaomi.jobs.f.mioffice.cn/index/position/7595847797652785446/detail)

## More jobs at Xiaomi 

### [27届招聘实习生-国际职能平台](https://jobs.opennash.com/companies/xiaomi/jobs/6d4d527313cd96474e53/)

北京

### [AI Agent 产品经理](https://jobs.opennash.com/companies/xiaomi/jobs/fc2f52865a7e2d84fbd7/)

北京

### [AI Agent 架构师 / Tech Lead](https://jobs.opennash.com/companies/xiaomi/jobs/35dcfe12c39db2a6a88a/)

北京

### [AI Agent产品经理](https://jobs.opennash.com/companies/xiaomi/jobs/1b2859d295836fa011a9/)

北京

### [AI Agent研究员实习生-小爱](https://jobs.opennash.com/companies/xiaomi/jobs/0c7ddc4025af5dc78885/)

北京

## Similar jobs at other companies

### [企业微信-大模型训练框架开发工程师-AI Infra（成都/北京）](https://jobs.opennash.com/companies/tencent-holdings/jobs/0dea41f9645c32af4070/)

Tencent Holdings · Guangzhou

OpenNash lists jobs from employers’ own careers sites and does not handle applications.
Dates come from the employer where stated; “first seen” and “last checked” are our own observations.

OpenNash Jobs · Official listings with dates and links to the employer. [Weekly hiring report](/reports/hiring/) · [Methodology](/methodology/)