Researcher
About Coase
Coase provides the infrastructure foundation for custom AI. By bringing together frontier training techniques, scalable infra, and proprietary models, we're driving the cost of intelligence ownership down to power the next generation of AI strategies. We're seeking team members that are extremely ambitious and want to make meaningful impacts on businesses at all scales.
Job Description
You will push the boundary on what's possible when non-experts fine-tune and deploy models. This is not a publish-papers-and-hope role. Every research outcome ships into a product that real customers use within weeks. You'll work at the intersection of efficient fine-tuning, inference optimization, and automated model evaluation.
Salary
$240,000 - $500,000 TC
Sample Projects
- Find new ways of generating large-scale, high-signal synthetic training data with coverage across important long-tail edge cases.
- Design novel ways to schedule post-training runs on limited GPU resources for maximal speed and cost efficiency.
- Write fused multi-LoRA attention kernels that batch requests across hundreds of different fine-tuned adapters sharing the same base weights.
Requirements
- Prior research in mathematics, statistics, physics, machine learning, or computer science.
- A need to move quickly and a desire to take on daunting problems.
- One example of a research project or paper with code you're proud of.
Bonus Points If
- You have worked on parameter-efficient fine-tuning, inference systems, or evaluation tooling that shipped into production.
- You are comfortable moving between research code, systems code, and product-facing constraints.
- You have experience optimizing workloads under real GPU, latency, or cost constraints rather than benchmark-only settings.