Member of Technical Staff - Partner Inference Engineer
## **技术人员成员 - 合作伙伴推理工程师** ### **关于液态 AI** 在 MIT CSAIL 中，我们构建了从数据中心加速器到设备上的硬件的通用人工智能系统，以有效地运行部署目标，确保低延迟，最小内存使用率，隐私和可靠性。我们与消费电子，汽车，生命科学和金融服务领域的企业合作。我们正在迅速扩展，需要特殊的人来帮助我们实现这一目标。### *机会**我们的论文是：液态基础模型可以解锁它们运行的的能力，包括 GPU,NPU 和 CPU. 许多玩家正在向我们运送硬件，以便 LFMs 可以在上面进行配置并将其带到市场上。今天没有人拥有这种功能。我们需要一个工程师能够把我们的模型降低到合作伙伴中，将它们配置，并将结果转化为合作伙伴及其客户可使用的证据。你将是此类公司的第一员工，并且将从头开始构建合作伙伴项目：设备实验室，个人资料制作过程和与合作伙伴工程团队的技术关系。大约70%的工作是实践技术；剩下的工作是面对合作伙伴和客户的，首先是汽车 OEM,下一步是 PCOEM. 你将每天与我们的 Edge Inference 团队合作，他们拥有你标记的运行时间修复。这个角色位于波士顿或旧金山。你拥有一台物理硬件，所以远程不是一个选择。### **我们正在寻找什么** 我们需要有人：- ** 了解：** 知道边缘 SoC 上的 NPU,GPU 和 CPU 块是如何组成的，模型如何降低到它们上，以及为什么芯片供应商想要一个模型伙伴。- ** 是这样做的，而不是计划的：** 个人已经把模型带到了边缘硬件上，测量延迟，功率和热力，并可以解释性能失去了哪里。- ** 想要面向合作伙伴的工作：** 可以在合作伙伴的客户工程师或 OEM 买家面前，共同创建演示和阅读他们的激励。如果你想仅仅在基本上工作，这是错误的角色，我们宁愿说现在。- ** 有模糊性：** 没有合作伙伴配置程序。你设计它，让像是这样工作，并手清诊断到边缘推理而不是围绕它们工作。### ** 工作 AI** - 所有和管理合作伙伴设备实验室：每一个 SoC Qualcomm,AMD,Intel,NVDA 以及其他芯片制造商为边缘和物理设备建造 SoCs 发送给我们的
查看英文原文
## **Member of Technical Staff - Partner Inference Engineer**
### **About LiquidAI**
Spun out of MIT CSAIL, we build general-purpose AI systems that run efficiently across deployment targets, from data center accelerators to on-device hardware, ensuring low latency, minimal memory usage, privacy, and reliability. We partner with enterprises across consumer electronics, automotive, life sciences, and financial services. We are scaling rapidly and need exceptional people to help us get there.
### **The Opportunity**
Our thesis is that Liquid Foundation Models unlock the capabilities of the silicon they run on, across GPU, NPU and CPU. Multiple silicon players are shipping hardware to us so LFMs can be profiled on it and taken to market together. Nobody owns that work today. We need an engineer who can lower our models onto partner silicon, profile them, and turn the results into evidence that partners and their customers can use.
You will be the first hire of this kind and will build the partner silicon program from scratch: the device lab, the profiling process, and the technical relationship with partner engineering teams. Roughly 70 percent of the role is hands-on technical; the rest is partner- and customer-facing, with automotive OEMs first and PC OEMs next. You will work daily with our Edge Inference team, who own the runtime fixes you flag.
This role is based in Boston or San Francisco. You own physical hardware, so remote is not an option.
### **What We're Looking For**
We need someone who:
- **Understands the silicon:** Knows how NPU, GPU and CPU blocks on an edge SoC are composed, how a model gets lowered onto them, and why a chip vendor wants a model partner.
- **Has done it, not planned it:** Has personally taken models onto edge hardware, measured latency, power and thermal, and can explain where performance was lost.
- **Wants the partner-facing work:** Can be put in front of a partner's customer engineers or an OEM buyer, co-create a demo, and read their incentives. If you want to work purely on fundamentals, this is the wrong role, and we would rather say so now.
- **Owns ambiguity:** There is no process for partner profiling. You design it, make the silicon work as it is, and hand clear diagnoses to Edge Inference rather than working around them.
### **The Work**
- Own and manage the partner device lab: every SoC that Qualcomm, AMD, Intel, NVDA as well as other chip makers building SoCs for edge and physical AI devices ship us
- Lower LFMs onto partner silicon and profile speed, latency, power and thermal
- Work directly with partner customer engineers to co-create demos and benchmark narratives
- Present results to partners and demand-side buyers, explaining why LFMs on their silicon beat the alternative
- Flag model and runtime performance gaps to Edge Inference with a clear hardware-level diagnosis they can act on
- Where capability allows, contribute to runtime and model-level optimization directly
### **Desired Experience**
Must-have:
- Deploying and profiling ML models on at least one edge SoC family (Qualcomm Snapdragon / Hexagon, AMD Ryzen AI / XDNA, Intel NPU, [SiMa.ai](http://SiMa.ai) MLSoC, or comparable)
- Working knowledge of at least one vendor lowering and runtime toolchain (Qualcomm AI Engine Direct / QNN, SiMa Palette, OpenVINO, ONNX Runtime execution providers, or similar)
- On-device performance measurement: latency, throughput, power and thermal
- Comfortable presenting technical results to external engineers and commercial stakeholders
Nice-to-have:
- Breadth across two or more chipset families
- Experience inside a chip vendor's customer-engineering, field-application or AI-enablement team
- Runtime or kernel-level optimization experience
### **What Success Looks Like (Year One)**
- LFMs are lowered and profiled on flagship chipsets from multiple silicon partners, with results packaged so the partner's own field teams can use them
- A working device lab and repeatable profiling process benchmarks new chipsets within weeks of arrival, not on Edge Inference's bandwidth
- Co-created demos are in active use in automotive and PC OEM conversations, and at least one qualified pipeline customer has come to Liquid through a silicon partner
### **What We Offer**
- **Impact at the core of the company:** This is where Liquid's thesis gets proven on real hardware, with the four biggest edge silicon names shipping chips to us to find out.
- **Compensation:** Competitive base salary with equity in a unicorn-stage company
- **Health:** We pay 100% of medical, dental, and vision premiums for employees and dependents
- **Financial:** 401(k) matching up to 4% of base pay
- **Time Off:** Unlimited PTO plus company-wide Refill Days throughout the year