Research Intern

Haize Labs builds safety tooling for LLMs  are the robustness layer eliminating the risk of using language models in any setting. To prevent these systems from failing, we preemptively discover all the ways in which they can fail and continuously eliminate them in deployment.

We are looking for Research Interns to help us develop fundamental safety tooling for LLMs. Your work will set the standard for research, testing, verification, and application of LLMs across customers, companies, and industries. You will directly influence how the world responsibly uses LLMs.

• Work directly with customers to adapt our core R&D for different domains.
• Build out core infra, cloud tooling, and UX around our algorithms.
• Deliver a delightful human-in-the-loop product experience.
• Ship tools that are used by developers across the world.

• Experience with ML in an applied setting.
• Strong open source presence or strong track record of software engineering projects and employment.
• Can ramp up very quickly on understanding our research.
• Love to break things, i.e. have a “stick it to the Man” attitude.

$125,000 – $200,000 USD

Location policy: 6 days a week, in person, in NYC.

US visa sponsorship: If you are exceptional, we will sponsor your visa.

We encourage you to apply even if you do not believe you meet every single qualification: We’re open to considering a wide range of perspectives and experiences, and would love to chat with you.

Compensation and Benefits: Haize Labs provides generous salary, equity, and benefits.

We are not here to write GPT wrappers or get rich quick off the AI bubble. We are here to work on the hardest, most fundamental research problem in AI: making it reliable and robust. Join us to push yourself, learn fast, experience excellence, and contribute to meaningful work. We value our team above all else and believe that greatness begets greatness.

Since starting 6 months ago, we have developed a suite of safety tools used by Anthropic, AI21, Scale AI, and other foundation model providers. We are backed by the founders of Cognition, Hugging Face, Weights & Biases, Nous, Etched, Okta, Replit, and others, and advised by professors from CMU and Harvard. Our founding team has a history of collaboration across notable roles and startups, including involvement with Stanford-affiliated programs and accelerators.