Chemical Week How can AI be scaled effectively in chemical manufacturing operations?
6 snips
Feb 25, 2026 
Dan Jeavons, president of Applied Computing and former Shell data science exec, builds foundation AI models for energy and petrochemical operations. He talks about tuning LLMs for industrial settings. He covers preventing hallucinations, combining physics and time-series data, top use cases like optimization and emissions, and the people and change-management challenges of adopting AI.
AI Snips
Chapters
Transcript
Episode notes
Combine Language Models With Physics For Plants
- Modern foundation models can integrate language abilities with physics and time-series to address whole-plant problems.
- Dan Jeavons explains Applied Computing combines LLM strengths with physics-informed and anti-hallucination layers for energy and petrochemicals.
Prevent Hallucinations By Constraining Models
- Constrain language outputs with physics, time-series data, and verification checks to prevent dangerous hallucinations.
- Jeavons describes training on public literature plus customer data inside the customer's environment and never exfiltrating those learnings.
One AI To Solve Multiple Plant Objectives
- Key operational objectives converge: optimization, reliability, integrity, safety, and emissions.
- Jeavons argues a single reconciled AI foundation can answer cross-discipline questions rather than many separate specialist tools.