Pinned
People, Technology & AI Research: My Blog and Projects 🤖
Hi, my name is Nick.
AI & Prompt Engineering in Research
I use UC Berkeley's Institute of Governmental Studies Library's collection of government documents to share techniques on using AI tools for research.
Latest articles
Teaching Gemini: How Models and Structured Prompts Actually Work
Google recently added a powerful new feature to Google AI Studio—it’s called Structured Prompts. You’ll find it in the dashboard when you’re working with Gemini, Google’s generative AI model. At first, it might sound confusing—Gemini, AI Studio, tuni...
When AI Meets HathiTrust: Why Old Catalogs Break OCR and What We Can Do About It
Most of the records in libraries like HathiTrust come from a time long before digital tools existed. Government reports, subject catalogs, and historical documents were created to be read by people—not machines. They were typed, printed, bound, and o...
Zooming Out: Why AI Projects Are Really About Automation
When it comes to AI, it’s easy to get lost in the sea of technical terms—metadata extraction, model training, prompt engineering, and more. But underneath all the jargon, one big idea drives almost everything: automation. Sure, I’ve been exploring ho...
🧠 Prompt Fine-Tuning Isn’t Enough: Why I Use Environmental Fine-Tuning
When you’re working with AI in real-world settings, writing good prompts is only half the battle. I say that because I’ve spent a lot of time testing and refining prompts to get more accurate, insightful results from AI. But here’s what I’ve learned:...
Testing the I-980 Workflow
Using AI to Discover Shifting Narratives, Sentiment & Visualization in One Place
Is AI Safe for Automating Workflows? Here’s What You Need to Know
Recently, someone asked me: “Is AI safe to use for automating information extraction, especially in Google Workspace?” My short answer: Yes—if you control how it’s built. When using AI with products like Google Colab, you’re operating within a truste...
Experimenting with AI for Metadata Trends: Training Models, Not Just Testing Extraction
AI is already capable of extracting metadata, identifying patterns, and analyzing trends over time—the real challenge isn’t proving that AI can do these things. Instead, the challenge is in training AI frameworks to do them efficiently, accurately, a...
How Do We Determine Research Relevance in AI-Generated Metadata?
📷 Image Reference: IGSL Document Example from Internet Archive Before diving into this post, you might find it helpful to read 3 Ways AI is Transforming Metadata for Large Scale Projects, where I explore AI’s role in extracting, generating, and conn...