Product & Program Leader · Engineer
Paul Davidson
Building for AI safety to mitigate the worst risks from Advanced AI.
I've spent 18+ years building hardware, software and experiences including data platforms that help people make high-stakes calls from complex, high-volume data. Most recently I worked on the experimentation platform at Expedia Group, the system teams used to evaluate changes before shipping them. My career started in electrical engineering and power substation design, and I'm bringing that foundation back to where I think I can make the largest impact now: ensuring we can mitigate the worst risks of advanced AI through supporting technical AI safety work.
- Tooling & Infrastructure
- AI Safety
- AI / LLMs
Currently
Refining my AI safety hypotheses and theory of change. Collaborating with a cohort of experienced builders that are mid-funnel in their AI Safety. Producing my first outputs within AI safety through my SPAR and a 2nd evals project. My latest AI Safety assessment.
My Story
I started out as an electrical engineer designing protection and control systems for power substations. Most of my career since has been in product: building support and documentation tools at National Instruments, launching their first chat products from scratch, and most recently running the internal experimentation platform at Expedia Group. The through-line has stayed the same. Take something ambiguous and high-stakes, give it structure, and help people make better decisions faster.
Two things shifted how I think about what to work on next. AI tools got good enough that a small, focused team can now build what used to take a department. At the same time, the problems I care most about, the safe development of AI and climate change, only got more urgent.
I'm now making a deliberate move toward that work. I initially focused on clean energy, but have now fully shifted my efforts into AI safety as the more urgent issue, and also a more direct transfer of my skills. I've worked through the base AI safety curriculum, am reading LessWrong and EA forums, making connections, starting to build outputs, and writing about what I learn along the way. This site is where it all lives. If you're building something in this space, let's connect.
Background
2022 – 2026
Senior Product Manager, Experimentation Platform
Expedia Group
Owned the internal A/B testing platform used across the company to evaluate product and ML changes before launch. Built automated rollout and rollback recommendations and real-time guardrails, and led the shift to measuring the platform by harm prevented rather than features shipped.
2011 – 2022
Product Manager: Chat, Documentation, Support Content
National Instruments
Eleven years across product and content roles. Launched ni.com's first chat products from scratch in six weeks, modernized the documentation platform, and built the function and governance for 6,000+ pages of technical support content.
2008 – 2011
Associate Engineer, Transmission Protection & Control
Santee Cooper
Designed protection and control systems for transmission substations end to end, from requirements through PE-stamped drawings and field validation, for a publicly owned utility serving coastal South Carolina.
B.S. Electrical Engineering
Clemson University · 2003 – 2007
AGI Strategy and Technical AI Safety - Certificate Courses
BlueDot Impact · June-August 2026
Right Now
Building
AI safety projects
SPAR project to centralize data collected during research to feed future research design loops. An evals project studying personas of judges.
This site
PaulsForge: somewhere to write, show projects, and think in public. Designed and built from scratch.
More coming soon
Lots of ideas queued, and ready to get them out for the world to see including a PoC for a end-end automated evals platform.
Exploring
AI safety and alignment orgs and operations · local LLMs and agent workflows · how small, AI-augmented teams ship faster
Writing About
What I'm learning in AI safety, lessons from building experimentation infrastructure at scale, how I'm using AI to accelerate my work, and notes from the shift into high-impact work.
Read the writings