机器学习工程师 — AI架构研究
Machine Learning Engineer — AI Architecture Research
职位描述
我们正在寻找一位专注于AI架构研究的机器学习工程师，以帮助设计、原型设计和验证下一代模型架构。你将在研究与生产之间工作——将新想法转化为可扩展的现实系统。
这个职位适合喜欢质疑架构假设、尝试新颖模型设计并超越标准Transformer方法的人。
你将参与的工作
- 研究和开发新的神经网络架构（例如Transformer的替代或扩展、循环/混合模型、长上下文系统）
- 设计并运行架构级别的实验（缩放定律、记忆机制、计算权衡）
- 从研究代码到训练就绪的实现，端到端地构建模型
- 与推理和系统工程师合作，确保架构可部署且高效
- 分析模型行为、失败模式和归纳偏见
- 阅读、复现并扩展前沿研究论文
- 为内部研究笔记、基准测试和开源项目（如适用）做出贡献
我们寻找的人选
- 扎实的机器学习基础和深度学习经验
- 有从零开始实现模型架构的实际经验
- 熟悉以下内容：
- 注意力机制、RNN、状态空间模型或混合架构
- 训练动态、缩放行为和优化
- 模型层面的内存、延迟和计算限制
- 熟练使用PyTorch或JAX
- 能够在理论、实验和工程之间自由切换
- 清晰的沟通能力，能够解释架构权衡
加分项
- 有非Transformer架构的经验（如RNN变种、SSMs、长上下文模型）
- 有以研究为导向的初创公司或开源ML项目的背景
- 有大规模训练或自定义训练循环的经验
- 有论文、预印本或显著的研究贡献
- 熟悉推理优化和部署约束
为什么加入我们
- 专注于核心模型架构，而不仅仅是微调
- 直接影响一家A轮公司的技术方向
- 小而高素质的团队，反馈速度快
- 有机会将研究成果投入生产
- 竞争性薪酬 + 有意义的股权
最初发布于Himalayas
查看英文原文
About the Role
We’re looking for a Machine Learning Engineer focused on AI architecture research to help design, prototype, and validate next-generation model architectures. You’ll work at the intersection of research and production — turning new ideas into scalable, real-world systems.
This role is ideal for someone who enjoys questioning architectural assumptions, experimenting with novel model designs, and pushing beyond standard Transformer-style approaches.
What You’ll Work On
- Research and develop new neural network architectures (e.g. alternatives or extensions to Transformers, recurrent / hybrid models, long-context systems)
- Design and run architecture-level experiments (scaling laws, memory mechanisms, compute trade-offs)
- Prototype models end-to-end — from research code to training-ready implementations
- Collaborate with inference and systems engineers to ensure architectures are deployable and efficient
- Analyze model behavior, failure modes, and inductive biases
- Read, reproduce, and extend cutting-edge research papers
- Contribute to internal research notes, benchmarks, and open-source efforts (where applicable)
What We’re Looking For
- Strong background in machine learning fundamentals and deep learning
- Hands-on experience implementing model architectures from scratch
- Solid understanding of:
- Attention mechanisms, RNNs, state-space models, or hybrid architectures
- Training dynamics, scaling behavior, and optimization
- Memory, latency, and compute constraints at the model level
- Comfortable working in PyTorch or JAX
- Ability to move fluidly between theory, experimentation, and engineering
- Clear communicator who can explain architectural trade-offs
Nice to Have
- Experience with non-Transformer architectures (RNN variants, SSMs, long-context models)
- Background in research-driven startups or open-source ML projects
- Experience with large-scale training or custom training loops
- Publications, preprints, or notable research contributions
- Familiarity with inference optimization and deployment constraints
Why Join
- Work on core model architecture, not just fine-tuning
- Direct influence on the technical direction of a Series-A company
- Small, high-caliber team with fast feedback loops
- Opportunity to ship research into production
- Competitive compensation + meaningful equity
Originally posted on Himalayas