LLM Research Scientist

GPP Database Link (https://cummins365.sharepoint.com/sites/CS38534/)

工作总结:

通过分析、数学和技术技能组合运用定量方法解决复杂分析问题。研究、设计、实施和验证运算法则，以分析多元化数据来源，通过利用复杂统计和预测建模概念实现针对性成果。

主要责任:

参与项目，通过使用数据科学方法支持主要目标和业务目标。

将数据科学方法用于更复杂的业务问题。

通过使用编程语言和工具运用统计方法创建多个运算法则。

与领域专家合作，验证模型能力。

与解决方案架构师合作实现适当数据流／数据模型、利用适当工具／技术的发展、快速原型和告知分析产品设计。

与经验不足的员工就数据科学工具和方法合作。

清楚地向利益相关者和同行群体阐明结果、方法和经验教训。

通过知识共享和协作，实现团队的持续发展和进步。

Qualifications

? JD #1 - LLM Research Scientist (Senior Level)

Cummins AI Lab - Artificial Intelligence Laboratory

About Cummins AI Lab

The Cummins AI Lab is a global innovation hub dedicated to advancing frontier AI capabilities and deploying AI solutions across Cummins' engineering, manufacturing, service, and industrial operations worldwide.

We focus on applied LLM research , domain-specific model development, and end-to-end enterprise AI transformation.

We are seeking a Senior LLM Research Scientist to drive advanced research, design next-generation domain LLMs, and influence the AI Lab's future technical direction.

Role Overview

This is a senior-level, high-autonomy role requiring deep expertise in large language models, applied AI research, and complex problem-solving. You will explore the latest LLM advancements, design LLM-driven solutions tailored to Cummins' engineering and industrial scenarios, and play a central role in building our domain-specific enterprise LLM.

Candidates must demonstrate strong applied research capabilities, cross-disciplinary thinking, and a proven ability to bring research into real-world production.

Key Responsibilities

1. Frontier LLM Research & Technology Exploration

• Investigate state-of-the-art LLM and multimodal technologies: RAG, Agents, tool use, reasoning, alignment, knowledge injection, and domain adaptation.
• Analyze global research trends and define how emerging techniques can be applied to Cummins' engineering and industrial workflows.
• Lead deep technical explorations and novel architecture experimentation.

2. Domain-Specific LLM Development

• Architect domain-grounded LLM solutions for engineering design, simulation workflows, diagnostics, service operations, manufacturing, and technical documentation.
• Build pipelines for SFT/IFT, retrieval-augmented reasoning, prompt engineering, and knowledge-grounded inference.
• Integrate structured and unstructured engineering knowledge (CAD/CAE outputs, diagnostic logs, manuals, sensor data).

3. Cross-Disciplinary Innovation

• Collaborate with engineering, simulation, materials, control systems, quality, and service teams to design innovative AI workflows.
• Apply multi-domain expertise to create LLM-powered tools for engineering productivity, automated reasoning, and smart decision-making.

4. End-to-End Research Delivery

• Own the full cycle: exploration → prototyping → evaluation → documentation.
• Produce technical reports, architecture designs, and executive-level presentations.
• Provide technical leadership in transitioning prototypes to enterprise systems.

Must-Have Qualifications

• Master's or PhD in Computer Science, AI, Machine Learning, Engineering, or related fields; OR equivalent senior-level research experience.
• Deep expertise in LLM technologies: Transformer, SFT/IFT, RAG, agents, tool-use, reasoning, alignment, evaluation.
• Proven experience shipping applied AI/LLM systems beyond PoC into real-world environments.
• Strong programming and experimentation skills (Python, PyTorch, HuggingFace, LangChain, LlamaIndex).
• Experience integrating LLMs with engineering, scientific, or industrial domains.
• Demonstrated ability to independently lead complex technical explorations.
• Strong communication skills for working with global research and engineering teams.

Preferred Qualifications

• Experience in AI Labs, cloud providers, or major technology companies.
• Experience building domain-specific or enterprise LLMs.
• Familiarity with CAD/CAE/PLM systems (Creo, SolidWorks, CATIA, Ansys).
• Publications, patents, or open-source contributions.

What You Will Gain

• Opportunity to shape Cummins' domain LLM strategy and long-term AI roadmap.
• Work on high-impact problems across engineering, manufacturing, and industrial operations.
• Collaboration with global experts across AI, product, engineering, and platform teams.

Responsibilities 技能:

协作 - 建立合作伙伴关系并与他人协作，以达成共同目标。

以顾客为中心 - 建立稳固的顾客关系，提供以顾客为中心的解决方案。

决策质量 - 及时作出高质量的决策，推动组织发展。

管理复杂情况 - 领会复杂的、大量的，有时甚至是相互矛盾的信息，以有效地解决问题。

精通技术发展 - 预见并采用有助于业务的信息技术和其他技术的创新。

数据挖掘 - 通过使用一系列数据探索和数据可视化技术来了解数据的底层结构，并在模型建立时得到良好结论，从而识别关系和模式，进而从数据提取洞见。

预测建模 - 通过使用适当的变量转换、特征选择策略、插补策略、类别重新平衡、重新取样策略和质量控制措施来开发分析模型或机器学习模型，以生成用于解决业务问题的预测性洞见。

程式设计 - 使用算法分析和设计、业界标准和工具、版本控制以及构建和测试自动化来创建、编写和测试计算机代码、测试脚本和构建脚本，以满足业务、技术、安全、治理和合规要求。

要求分析 - 根据要求的复杂性和对企业的价值评估要求之间的关系和相互依赖关系，以确定可行性和优先顺序。

统计建模 - 利用零假设显著性检验、回归模型、广义线性模型、时间序列分析、秩统计量、概率分布拟合生存分析等基础统计学知识，建立回归、分类、孤立点检测、异常检测、时间序列预测的统计解释与描述模型，从而为任何给定的统计或业务问题验证假设。

解决问题 - 通过利用行业标准方法来创建问题可追溯性并保护客户，使用系统分析流程解决问题，并可以指导他人有效解决问题；确定可指定的原因；实施完善、基于数据的解决方案；识别系统性根源并采取建议措施防止问题再次发生。

看重差异性 - 认识到不同视角和文化给组织带来的价值。

教育,资格,认证:

要求具有相关技术学科的大专、本科或同等学历，或相关的同等工作经验。该职位可能需要获得有关遵守出口管制或制裁法规的许可证。

经历:

需要具有相关学科领域的中等经验，并具有分析复杂业务系统和大型数据集的可证明的记录。了解最新的数据科学技术和趋势者更加优先，包括:

• 熟悉分析复杂的业务系统、行业需求和/或数据规则
• 处理和管理大型数据集的背景
• 大数据、开源、第三方工具集应用知识
• SQL 查询语言
• 基于云的集群计算实施经验
• 建立分析解决方案的经验

有以下领域中等经验者优先:

• 使用开源和第三方工具来实施大数据平台解决方案
• Microsoft Azure 和/或 Amazon Web 服务环境
• 敏捷软件开发的经验
• 熟悉机器学习系统的验证和测试
• 熟悉持续集成和持续交付 (CI/CD)

Job Systems/Information Technology

Organization Cummins Inc.

Role Category Off-site Remote

Job Type Exempt - Experienced

ReqID 2424134

Relocation Package No

100% On-Site No

Client-provided location(s): Beijing, China

Job ID: Cummins-R-74535985196347F3BE412E7AA25091A7

Employment Type: OTHER

Posted: 2026-02-28T19:24:42