Featured on Nexus Labs: A Podcast About Autonomous Building Control

Written by
Danielle Radden
Published on
May 13, 2025

Hey fellow rebels.

Recently facil.ai was highlighted in a Nexus Podcast episode. It features our customer Kenny Seeton from California State University, Dominguez Hills (CSU-DH) and our own CEO Keith Gipson. They share an incredible story using AI to revolutionize central plant operations. Seriously, the results are mind-blowing. I've distilled the conversation down to the top 10 takeaways, focusing on the real-world challenges, solutions, and lessons that we can all relate to.

Listen Here

Project Summary and Takeaways

  • CSU Dominguez Hills aims to be the first net zero CSU campus, leading the system by adopting innovative technologies to reduce greenhouse gas emissions and improve energy efficiency.
  • The central plant at CSU-DH serves about 1.5 million square feet across 13 main buildings, all connected to a single plant via a tunnel system, enabling efficient monitoring and maintenance.
  • The campus transitioned from gas-fired absorption chillers to three 1,000 ton electric, magnetic-bearing chillers and now operates the largest heat pump project on the west coast, reducing gas use by 95%.
  • The chiller optimization project was a close collaboration between CSU-DH and facil.ai, with a small internal team and deep vendor involvement. The campus continues to serve as an R&D lab for the technology.
  • Facil.ai’s autonomous optimization system directly writes setpoints to the chiller plant every five minutes, using deep reinforcement learning to continually improve efficiency without operator intervention.
  • The AI-agents optimize the entire plant—not just the chiller—by balancing the energy use of chillers, pumps, and cooling towers, finding the “sweet spot” for total kW/ton rather than just focusing on the coldest water possible.
  • After implementation, the plant’s average kW/ton dropped from 0.86 in 2022 to 0.45 in 2024, representing a roughly 48% improvement in energy efficiency for the whole system.
  • The AI system is designed with safety guardrails, including a kill switch, so if the AI goes offline or malfunctions, the system reverts to its previous control logic, ensuring operational safety.
  • The approach is highly repeatable and scalable: facil.ai has begun deploying the solution at other large central plants, and the technology does not require extensive on-site vendor presence.
  • Lessons learned include the importance of building trust with operations teams, implementing robust fail-safes, and recognizing that AI-agent optimization is an “add-on” that doesn’t disrupt existing workflows but delivers significant efficiency gains.

Conclusion

Let's face it: we're all under pressure to do more with less, and the CSU-DH and facil.ai partnership offers a compelling example of how to achieve that. By embracing innovative AI technology and prioritizing safety we've not only slashed energy consumption but also created a more resilient and sustainable operation. As operators of our built world, we're always looking for practical solutions that deliver tangible results. This project proves that AI isn't just hype; it's a powerful tool that can help us meet our goals and create a better future for our communities.

Listen to the Full Podcast

Easy? Sí, Facil. Easy? Sí, Facil. Easy? Sí,

Easy? Sí, Facil. Easy? Sí, Facil. Easy? Sí,