机器学习平台与推理工程师 - 内容开发工程师
ML Platform and Inference Engineer - Content Developer
Jala University 是一项创新计划，旨在通过提供符合行业需求的实践教育，弥合学术界与产业之间的差距，采用独特的教育模式，整合来自学术界和产业界的专家。
我们的目标是通过软件行业改变未充分服务地区的经济，创造对个人、社区和地区产生影响的职业机会，同时为后代留下持久的遗产。
职位要求
我们正在寻找一名内容开发者，为人工智能系统工程硕士课程设计课程内容、实验、参考实现、评估工具包和教师指南。
职位要求
- 5年以上交付生产级软件的经验，其中2年以上在生产环境中的AI/ML经验
- 能够亲自将课程素材构建到生产标准
- 有公开的技术写作证据（文档、技术博客、开源项目）
- 具备可复现性纪律（依赖项固定、容器化、种子运行）
- 具备专业的英文书面表达能力
- 愿意将开发的材料交给项目（课程将由其他人教授）
- 在微调经典机器学习模型/小型基础模型方面有记录
- 有将容器化推理服务交付生产的经验
- 有在生产环境中使用多租户服务的经验（路由、队列、隔离）
- 在量化和缓存方面有记录，并能证明服务成本的降低
技术栈
Docker, Modal (CPU/T4/A10和H100/A100), Hugging Face Hub, scikit-learn, PyTorch, vLLM, Prometheus, Grafana, Locust 或 k6
福利
- 远程工作模式（居家办公）。
- 加入一家充满活力且不断发展的具有国际影响力的组织。
最初发布于 喜马拉雅山
查看英文原文
Jala University is an innovative initiative designed to bridge the gap between academia and industry by delivering practical education tailored to industry needs, with a unique educational model integrating experts from both academia and industry.
Our goal is to transform the economies of underserved regions through the software industry, creating professional opportunities that impact individuals, communities, and regions, while leaving a lasting legacy for future generations.
Requirements
We're looking for a content developers for its Master in AI Systems Engineering. You'll design module content, labs, a reference implementation, an evaluation harness, and an instructor guide.
Requirements
- 5+ years shipping production software, including 2+ in production AI/ML
- Able to personally build the course artifact to production standard
- Public evidence of technical writing (docs, technical blog, open-source project)
- Reproducibility discipline (pinned dependencies, containers, seeded runs)
- Professional written English
- Willing to hand over developed material to the program (course will be taught by others)
- Track record fine-tuning classical ML / small foundation models
- Experience shipping containerized inference services to production
- Experience with multi-tenant serving in production (routing, queueing, isolation)
- Track record in quantization and caching, with demonstrable serving cost reduction
Tech Stack
Docker, Modal (CPU/T4/A10 and H100/A100), Hugging Face Hub, scikit-learn, PyTorch, vLLM, Prometheus, Grafana, Locust or k6
Benefits
- Remote work modality (home office).
- Joining a dynamic and growing organization with international reach.
Originally posted on Himalayas