- San Francisco, California, United States
- 5–10 years
Extend what is possible with Lumafield’s agentic tools by working across the agent stack, from top level harness and eval development down to fine-tuning and pre-training of foundational encoders for spatial reasoning.
Compensation
Lumafield offers both competitive cash and equity compensation, as well as a health & wellness stipend, 401k, parental leave, flexible PTO, commuter benefits, company wide events and more!
Responsibilities
- Benchmark local LLMs, VLMs, and perception encoders
- Fine tune and post-train models on domain-specific data and tasks
- Develop harnesses and build real, deployed agents that our customers use
- Develop evaluations that predict real world agent performance accurately
- Partner with product and engineering to continually improve agents and agentic workflows
Requirements
- Advanced degree in machine learning, optimization, or a relevant discipline
- Nice to have: design
- Nice to have: manufacturing
- Nice to have: hardware
- Nice to have: 2D or 3D data