About
the workshop.
Brian Hengen — VP at Oracle, AI researcher, and builder of specialized language models.
The Mission
This blog documents my hands-on exploration of Specialized Language Models (SLMs)—the hypothesis that smaller, domain-focused AI models can outperform general-purpose giants on specific tasks.
Since September 2025, I’ve been building, fine-tuning, and testing this thesis with real applications:
| Project | What I Learned |
|---|---|
| ChefBot | A cooking AI that coordinates multi-dish meal prep—specialists excel at domain knowledge |
| Executive Assistant | An AI trained on 5+ years of my management notes—data quality trumps model size |
| 3-LLM Toggle | Balancing cloud, local, and GPU inference—flexibility beats optimization |
| Multi-Model Collaboration | Two AIs pair programming—specialists need generalists, and vice versa |
Every experiment runs on my RTX 5090 workstation, with findings documented here for others walking a similar path.
About Me
I’m Brian Hengen, Vice President at Oracle, where I lead technical sales engineering teams focused on AI and database technologies.
By day, I help enterprises adopt AI-powered data solutions. By night (and early mornings), I build things—fine-tuning models, writing code, and testing ideas that might not work.
What drives me:
- The belief that AI should be practical, not theoretical
- A conviction that smaller, specialized models have an outsized role to play
- The joy of building something that actually runs
Background:
- 20+ years in technology leadership
- Self-taught programmer since age 7
- Former Oracle DBA who learned by being thrown into the deep end
- No college degree—just relentless curiosity
Connect
- GitHub: github.com/bjhengen
- LinkedIn: linkedin.com/in/brian-hengen
- X/Twitter: @bhengen
Questions about the experiments? Found a bug in my code? Want to collaborate? Reach out.
Disclaimer
The views expressed here are my own and do not represent my employer. I don’t disclose confidential information or use corporate branding. This is a personal space for learning and sharing.