Impactful AI Research Engineer: Build ‑ Systems

About LG AI Research Center, Ann Arbor

The LG AI Research Center in Ann Arbor, inaugurated in March 2022, focuses on cutting‑edge AI research with the goal of developing impactful and responsible AI, with a mission to propel technological innovations, advance scientific discovery, and benefit all of humanity  foster an environment that encourages open communication, collaboration, diverse perspectives, and a growth mindset. Whether you are a well‑established expert or a high‑potential candidate eager to dive into topics that resonate with our mission, we welcome your contribution to our team.

We seek individuals who are excellent not just in technical ability, but in their alignment with our values and the meaningful impact of their work. If you are driven by making a difference and desire your work to resonate beyond academic publications, join us in shaping the future of AI.

We do not discriminate against candidates based on nationality, sex, age, religion, disability, or other legally protected statuses.

• Develop and implement engineering‑driven research ideas with a high degree of autonomy and initiative, ensuring these ideas are both innovative and practical.
• Align research efforts with the needs of both internal and external users, ensuring that research products are scalable and demonstrate practical utility in varied applications.
• Collaborate on impactful research projects that address complex, real‑world problems, focusing on delivering tangible results.
• Contribute to creating demos or systems that highlight the team's research and engineering efforts, demonstrating our AI innovations' practical applications and real‑world effectiveness while fostering interdisciplinary collaboration with other teams.
• Innovate and improve state‑of‑the‑art models, datasets, architectures, and algorithms within machine learning.
• (optional) Publish in top‑tier conferences and journals and effectively communicate our AI research findings to both experts and broader audiences.

Currently, we have two focus areas in research:

1. Develop AI agents capable of helping human users by solving tasks in complex environments, harnessing the latest advances in foundational models (such as large language models and large multimodal models), sequential decision‑making, reinforcement learning, and multimodal AI.

2. Advance research and development of foundation models (e.g., large language models, multimodal foundation models, etc.) by developing compute‑efficient training algorithms and architectures, constructing large‑scale, high‑quality datasets automatically, and improving generalization and adaptability with transfer learning and few‑shot/zero‑shot learning.

• Large Language Models
• Compute‑efficient training algorithms and architectures
• Curating and building large‑scale high‑quality datasets/benchmarks automatically
• Improving instruction following and text generation via advanced instruction tuning and RLHF
• Reasoning
• Code generation
• LLM Agents
• Dialogue Systems

• RL + Language
• LLM Agents
• Compositional task generalization
• AI agents/assistants for real‑world applications
• Hierarchical reinforcement learning/planning/imitation learning
• Meta/multi‑task/transfer reinforcement learning
• Reinforcement learning from human feedback (RLHF)
• Offline reinforcement learning

• Vision‑language grounding
• Video understanding
• Deep generative models (images, videos, text, etc.)

• Strong programming skills and an impressive project portfolio, with a focus on producing work that has real‑world applicability.
• Proficient in the use of deep learning frameworks, with strong skills in applying these tools to solve complex problems.
• Expertise in state‑of‑the‑art AI research topics and methodologies, demonstrating a thorough understanding of the field's current landscape and its future potential.
• Excellent communication skills, proficient at clearly presenting the research vision, technical concepts and methodologies, experimental results, and key insights to a diverse audience, with the ability to tailor the message for varying purposes and levels of technical…