# Nick Joseph

Research and product · Head of Pre-training at Anthropic · joined 2021

Joined OpenAI for its safety team, then followed its safety leads out eight months later. He has been at Anthropic since January 2021.

Full profile: https://www.autumn.ai/resources/anthropic/people/nick-joseph

> “after, like, eight months, basically everyone I worked with—all the safety leads—left, which, uh, yeah, invited me to go to Anthropic. And that was sort of the reason I joined OpenAI in the first place—because I cared about AI safety and wanted to work with them. So then I went with them to join Anthropic pretty much right when it started.” [1](https://www.ycombinator.com/library/Mw-how-to-train-an-llm-with-anthropic-s-head-of-pretraining)

**At Anthropic:**
- Current: Head of Pre-training
- Head of Training
- Jan 2021: Member of Technical Staff
Sources: [1](https://www.ycombinator.com/library/Mw-how-to-train-an-llm-with-anthropic-s-head-of-pretraining) [2](https://podcasts.apple.com/us/podcast/197-nick-joseph-on-whether-anthropics-ai-safety-policy/id1245002988)

**Career before Anthropic:**
- May 2020 – Jan 2021: OpenAI, Member of Technical Staff
- Dec 2017 – May 2020: Vicarious AI, Research Engineer
Sources: [1](https://www.linkedin.com/in/njoseph2)

**Education:**
- Bachelor's degree, Computer Science, Harvard University, 2017
Sources: [1](https://www.linkedin.com/in/njoseph2)

**Lesser-known facts:**
- Interned at GiveWell and first planned a career fighting poverty. [1](https://www.ycombinator.com/library/Mw-how-to-train-an-llm-with-anthropic-s-head-of-pretraining)
- First job: training computer-vision models for robots at Vicarious. [1](https://www.ycombinator.com/library/Mw-how-to-train-an-llm-with-anthropic-s-head-of-pretraining)
- Was one of the 11 people who left OpenAI to launch Anthropic. [1](https://80000hours.org/podcast/episodes/nick-joseph-anthropic-safety-approach-responsible-scaling/)
- Studied computer science at Harvard. [1](https://www.linkedin.com/in/njoseph2)
- Worked on scaling laws for code generation on OpenAI's foresight team. [1](https://www.youtube.com/watch?v=qscouq3lo0s)
- Argues the bottlenecks in AI progress are people and time, not data or compute. [1](https://80000hours.org/podcast/episodes/nick-joseph-anthropic-safety-approach-responsible-scaling/)
- Says Anthropic's responsible scaling policy avoids costly precautions before they are needed and aligns commercial incentives with safety. [1](https://80000hours.org/podcast/episodes/nick-joseph-anthropic-safety-approach-responsible-scaling/)
- Welcomed Andrej Karpathy to his pretraining team in May 2026 to build a group using Claude to accelerate pretraining research. [1](https://x.com/nickevanjoseph/status/2056760504949842219)

**Interviews and talks:**
- [80,000 Hours podcast on Anthropic's safety approach and responsible scaling (2024)](https://80000hours.org/podcast/episodes/nick-joseph-anthropic-safety-approach-responsible-scaling/)
- [Y Combinator: How to train an LLM with Anthropic's head of pretraining (2025)](https://www.ycombinator.com/library/Mw-how-to-train-an-llm-with-anthropic-s-head-of-pretraining)
- [PyTorch talk: Why and how of scaling large language models (2022)](https://www.youtube.com/watch?v=qscouq3lo0s)

Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-10-18. Everyone at Anthropic: https://www.autumn.ai/resources/anthropic/people