Understanding AI. Engineering what matters.
I’m VP of Engineering & Technology at Bear Creek AI, where I help build knowledgeXpert. I research and explore how AI produces its answers, where it fails, and which work is worth handing to it.
Current work
VP of Engineering & Technology
- I set the technical direction and architecture for our AI software, which serves engineers and technical organizations.
- I hire and lead the software team that builds it.
- I test new models against real engineering work before they go into the product.
Selected project
knowledgeXpert
- The problem
- Engineering firms sit on hundreds of thousands of documents: standards, project reports, technical notes. I watched teams stop critical work to hunt for one spec line.
- My contribution
- I joined as a founding engineer and helped build the product. I now lead its architecture and the team behind it.
- What it helps engineers do
-
- Ask an engineering question and get a concise answer that points to the source document.
- Organize PDFs, spreadsheets, images and Word files into secure, searchable knowledge bases.
- Search published industry libraries alongside their own files.
Writing
-
How much protection does “human in the loop” give you if the human approves everything?
OWASP’s Top 10 for Agentic AI names Human-Agent Trust Exploitation. Asking for approval at every step trains people to approve automatically. Which agent actions deserve a real review?
-
Confidence vs. Competence
We read fluent, confident answers as accurate ones. LLMs produce the signals of expertise whether or not the knowledge is there.
-
The Age of Infinite Drafts
AI made starting cheap. It did nothing to tell you what is worth finishing. Speed is not the same thing as velocity.
-
Most people don’t understand their daily workflows
Can you write down the repeatable steps in your own week? Until you can, AI stays stuck on single-step tasks.
Speaking
Introduction to AI for Engineering
Generative AI Best Practices in Engineering
I speak to engineers and business leaders. The questions I cover:
- How does a language model produce an answer, and why can it be confidently wrong?
- What does a benchmark score tell you about a model, and what doesn’t it?
- Which of your work can you hand to AI, and what do you have to define first?
- When does a human approval step catch a mistake, and when is it a rubber stamp?
Background
- Education
- Bachelor’s in Electrical/Computer Engineering
- Engineering licensure
- FE Computer Engineering
- Certificates
-
- Security Engineering for Agentic AI Systems
- Fundamentals of Machine Learning and Artificial Intelligence
- Introduction to Cloud Computing
- DevOps Essentials
- Where it started
- In high school I ran an Etsy shop selling LED prop lightsabers, took custom 3D printing orders, and taught video game and coding courses. The shop and the print jobs taught me to build what the customer asked for instead of what I assumed they wanted. The courses taught me to explain a technical idea to someone meeting it for the first time, and I have been teaching in some form since.
Contact
For talks, workshops or professional collaboration, email me.
- Emailcgdalton01@gmail.com
- LinkedInClayton Dalton on LinkedIn
- CompanyBear Creek AI