高级研究工程师，大语言模型训练与后训练
Senior Research Engineer, LLM Training & Post-Training
### 我们是谁
Lightning AI 是 PyTorch Lightning 的开发公司。成立于 2019 年，我们构建了一个端到端的平台，用于开发、训练和部署 AI 系统——旨在减少摩擦，将想法从研究阶段快速推进到生产环境。
通过与 Voltage Park（一家新云和 AI 工厂）的合并，Lightning AI 将以开发者为中心的软件与高效、大规模的计算相结合。团队可以获得实验、训练和生产推理所需的工具，同时内置安全、可观测性和控制能力。
我们服务独立研究人员、初创企业和大型企业。Lightning AI 全球运营，在纽约市、旧金山、西雅图和伦敦设有办公室，并获得 Coatue、Index Ventures、Bain Capital Ventures 和 Firstminute 的支持。
### 我们的工作方式
在这里茁壮成长的人是快速行动者，他们公开沟通、承担责任，并持续提升自己、团队和公司。以下是具体表现：
- **快速行动：** 我们快速推进，做出深思熟虑的决策，保持势头。我们重视行动胜于完美，通过发布产品来学习。
- **承担责任：** 我们对结果负责，而不仅仅是个人工作。我们做出推动公司前进的决策并坚持到底。
- **公开沟通：** 我们直接沟通，寻求理解，并为他人创造清晰度。诚实的对话帮助我们更快地一起前进。
- **打造优秀团队：** 我们以身作则，赋能他人，并创建健康团队，让员工能够发挥最佳水平。
- **提高标准：** 我们不断自我提升。我们从反馈中学习，持续挑战自己成长，并专注于最重要的工作。
- **长远思考：** 我们为未来设计。我们创建可扩展的系统，简化复杂性，并利用 AI 和自动化来放大我们的影响力。
### 我们寻找的人选
我们正在寻找一位经验丰富的 **高级研究工程师**，有构建、训练和优化现代基于 transformer 的语言模型的经验，加入 Lightning AI 的研究工程团队。
该职位将专注于改进大型语言模型在 Lightning AI 平台和实际客户工作负载中的训练、微调、评估和部署方式。该职位将跨模型训练、后训练、PyTorch、分布式系统和 AI 系统工程工作，以提高模型质量、训练效率和开发人员生产力，同时与研究人员紧密合作。
查看英文原文
### Who **We Are**
Lightning AI is the company behind PyTorch Lightning. Founded in 2019, we build an end-to-end platform for developing, training, and deploying AI systems—designed to take ideas from research to production with less friction.
Through our merger with Voltage Park, a neocloud and AI Factory, Lightning AI combines developer-first software with cost-efficient, large-scale compute. Teams get the tools they need for experimentation, training, and production inference, with security, observability, and control built in.
We serve solo researchers, startups, and large enterprises. Lightning AI operates globally with offices in New York City, San Francisco, Seattle, and London, and is backed by Coatue, Index Ventures, Bain Capital Ventures, and Firstminute.
### The Way We Work
The people who thrive here are builders who move fast, communicate openly, take ownership, and continuously improve themselves, their teams, and our company. Here's what that looks like in practice:
- **Move with Urgency:** We move quickly, make thoughtful decisions, and keep momentum. We value action over perfection and learn by shipping.
- **Take Ownership:** We own outcomes, not just our individual work. We make decisions that move the company forward and follow through.
- **Communicate Openly:** We communicate directly, seek to understand, and create clarity for others. Honest conversations help us move faster together.
- **Build Great Teams:** We lead by example, empower others, and create healthy teams where people can do their best work.
- **Raise the Bar:** We're always improving ourselves. We learn from feedback, consistently challenge ourselves to grow, and focus on the work that matters most.
- **Think Long-Term:** We design for what's next. We create scalable systems, simplify complexity, and use AI and automation to amplify our impact.
### What We're Looking For
We're are looking for an experienced **Senior Research Engineer** who has built, trained, and optimized modern transformer-based language models to join our Research Engineering function at Lightning.
This role will focus on advancing how large language models are trained, fine-tuned, evaluated, and deployed across Lightning AI's platform and real-world customer workloads. It will work across model training, post-training, PyTorch, distributed systems, and AI systems engineering to improve model quality, training efficiency, and developer productivity while collaborating closely with researchers, infrastructure engineers, and customers.
We're looking for someone who enjoys turning cutting-edge research into production systems. You have deep experience training and improving transformer-based language models, strong software engineering fundamentals, and a passion for solving difficult problems across model training, evaluation, and AI systems. Rather than building applications on top of existing models, you're motivated by improving the models themselves and the systems that power them. Our work spans models that power the Lightning AI platform, customer-specific model workloads, and research that translates into reusable training and platform capabilities.
_This role is hybrid with a minimum of 2 in-office days per week in San Francisco, Seattle, NYC, or London, with fully remote work considered for candidates outside of our office hub locations. All employees participate in occasional team and company offsites._
### **What You'll Do**
- Design, build, and optimize training and post-training pipelines for large language models.
- Improve model quality through supervised fine-tuning, continued pretraining, preference optimization, reinforcement learning, evaluation, and experimentation.
- Build and improve PyTorch-based training infrastructure, tooling, and developer workflows.
- Optimize distributed training across multi-GPU environments by improving throughput, memory efficiency, scalability, and GPU utilization.
- Investigate model training issues, including convergence, instability, communication overhead, and performance bottlenecks.
- Design evaluation methodologies, benchmark models, analyze failure modes, and acheive model improvements through experimentation.
- Collaborate directly with customers to understand real-world workloads and translate those learnings into improvements across Lightning AI's research platform.
- Partner closely with research, infrastructure, and platform engineering teams to build production-ready AI systems.
- Contribute to open-source projects through new features, tooling improvements, documentation, and community engagement
### **What You’ll Need**
#### **Required Qualifications**
- Significant experience training, fine-tuning, evaluating, and/or optimizing transformer-based language models using PyTorch.
- Experience with modern LLM training and post-training techniques such as continued pretraining, SFT, RLHF, preference optimization (DPO, PPO, GRPO), reward modeling, or similar approaches.
- Strong understanding of distributed training and multi-node systems, with experience improving training performance, scalability, and/or efficiency.
- Strong software engineering fundamentals, including building production-quality Python software and research tooling.
- Experience designing experiments, evaluating model performance, and debugging complex training and/or optimization issues.
- Excellent communication and collaboration skills, including the ability to work effectively across research, product, infrastructure, and customer-facing engagements.
- Comfortable working in fast-moving, ambiguous environments where priorities evolve over time.
- Master's degree, PhD, or equivalent industry experience in Machine Learning, AI, Computer Science, or a related field
#### **Ideal Experience**
Experience with one or more of the following:
- DeepSpeed, FSDP, Megatron-LM, Hugging Face Transformers, TRL, PEFT, Lightning Fabric, NVIDIA Molt, or similar training frameworks.
- CUDA, Triton, vLLM, SGLang, TensorRT, or other AI systems and performance optimization technologies.
- GPU performance optimization, mixed precision, memory optimization, or distributed training optimization.
- Open-source contributions, research publications, or production AI platforms supporting large-scale training or inference workloads.
- Startup experience or experience working on highly cross-functional engineering teams.
### Compensation
We are committed to offering competitive compensation that reflects the value each team member brings to our mission. Final offers are based on factors such as experience, skills, geographic location, and role expectations. In addition to base salary, our total rewards package for eligible roles includes a discretionary bonus, a meaningful equity component, and comprehensive benefits.
The anticipated annual base salary range for this role is:
$165,000—$310,000 USD
### **Benefits and Perks**
We offer a comprehensive and competitive benefits package designed to support our employees’ health, well-being, and long-term success:
- **Comprehensive Health Coverage:** Medical, dental, and vision coverage for employees and eligible dependents.
- **Meaningful Equity:** RSUs that give employees a stake in the company's long-term success.
- **Retirement Savings:** 401(k) matching (U.S.) and pension contributions (U.K.).
- **Flexible Time Off:** Unlimited PTO, company holidays, and floating holidays to support work-life balance.
- **Company-Wide Winter Break:** Two weeks of company closure each winter to disconnect and recharge.
- **Paid Parental & Family Leave:** Paid leave to support you and your family through life's important moments.
- **Professional Development:** Annual learning and development allowance to support your professional growth.
- **Wellness Benefits:** Wellness and work-from-home stipends to support your physical and mental well-being.
- **Sabbatical Program:** Four weeks of paid sabbatical leave after four years of service.
- **Flexible Work:** Flexible schedules and a hybrid work model for our office-based teams.
- **In-Office Meals:** Complimentary meals at our office hubs.
Benefits may vary by location, team, and role.
_At Lightning AI, we are committed to fostering an inclusive and diverse workplace. We believe that diverse teams drive innovation and create better products. We provide equal employment opportunities to all employees and applicants without regard to race, color, religion, gender, sexual orientation, gender identity, national origin, age, disability, veteran status, or any other protected characteristic. We are dedicated to building a culture where everyone can thrive and contribute to their fullest potential._