LLM Algorithm Engineer

About the Role

Enhance large language models (LLMs) and agent functionalities via dedicated post-training experimentation and research. Collaborate across data management, training techniques, and evaluation methods to identify performance bottlenecks and integrate research advancements into practical applications.

Responsibilities

• Conduct research and improve LLM post-training techniques and training protocols, creating experiments focused on specific tasks to boost effectiveness and generalization.
• Develop and refine training datasets, optimizing quality, spread, and application strategies aligned with capability objectives and experimental feedback.
• Create benchmarks and testing frameworks that reflect key capabilities and real-world scenarios, establishing reliable measures for comparing training methodologies.
• Examine model behaviors and persistent errors, propose potential improvements, validate through training trials, and document outcomes comprehensively.
• Collaborate with product development, engineering, and modeling partners to integrate application insights into data, training, and assessment processes, evaluating impact on actual tasks.

Requirements

• Strong background in machine learning and large language models, with scholarly or hands-on experience in large-scale foundation model training or fine-tuning, including knowledge of supervised fine-tuning (SFT), reinforcement learning, or preference optimization.
• Expertise in popular deep learning frameworks, capable of independently designing experiments, interpreting training data, and troubleshooting issues.
• Comprehensive understanding of the influence of data and training strategies on model outcomes, with proficiency in designing enhancements informed by evaluations and user scenarios.
• Active participation in relevant academic research with sound judgment to generate testable hypotheses from open-ended challenges and conduct thorough evaluations.

About Manus AI

Manus AI delivers outstanding productivity solutions that empower individuals to accomplish tasks effortlessly.

What We Offer

• Opportunity to develop at the forefront of AI agents, collaborating with an agile team transforming advanced AI breakthroughs into meaningful applications.
• Equity participation through a share incentive program, allowing team members to benefit from long-term company growth and shared value creation.
• Access to unlimited Manus tokens to enable experimentation, creation, and enhancement of individual productivity.