LLM Algorithm Engineer

Menlo Park, CA
· 5 days/week in office
· Full-time

We operate at the intersection of frontier model research and production-grade applied AI. Our customers are not looking for demos. They need agents that drive revenue, support underwriting decisions, and perform reliably in real business environments.

We are early, well-resourced, and moving fast.

We are looking for a Large Language Model Algorithm Engineer to help build, fine-tune, align, optimize, and deploy large language models for real-world AI systems. This role focuses on model training, post-training, domain-specific model development, evaluation, inference optimization, and integration into agent systems.

• Own fine-tuning, post-training, and performance optimization for large language models.
• Build domain-specific LLMs, including data design, training, and evaluation.
• Lead model alignment efforts, including SFT, DPO, RLHF, and related data strategy optimization.
• Explore reinforcement learning applications in large language models, including PPO, actor-critic methods, and related approaches.
• Build an end-to-end model iteration loop, covering data, training, evaluation, and continuous improvement.
• Contribute to model inference optimization, including distillation, quantization, acceleration, and deployment.
• Support the practical deployment of models in agent-based systems.

• Master’s degree or above in computer science, mathematics, artificial intelligence, or a related field.
• Strong foundation in deep learning and reinforcement learning.
• Solid understanding of Transformer architecture and large-scale model training workflows.
• Experience with LLM fine-tuning or post-training, such as SFT, DPO, or RLHF, is preferred.
• Proficient in PyTorch and familiar with Linux and GPU-based development environments.
• Strong modeling ability and engineering implementation skills.
• Strong sense of ownership and a continuous improvement mindset.

• Foundational work:
  Join an early team building both model infrastructure and applied AI products.
• Real product focus:
  We train our own models and build agents designed for production use, not demos.
• High-impact problems:
  Our agents are built for business-critical workflows such as lead generation and underwriting.
• Fast execution:
  We are a tight team with low bureaucracy and a strong bias toward shipping.
• In-person culture:
  We work from our SF Bay Area office five days a week.

Corepass is an equal opportunity employer. We hire on merit and welcome candidates of every background.