AI for EE Research Productivity¶

About Dan McCreary¶

BA in Physics/Computer Science from Carleton College in 1982
MSEE in Electrical Engineering/Computer Science from the University of Minnesota in 1983
DSP design at Bell Laboratories
Supercomputing software (Unix)
Worked for Steve Jobs at NeXT Computer
Founded Integrity Solutions (software services)
Data exchange standards for State of Minnesota
NoSQL Conferences and NoSQL textbook
Distinguished Engineer at UGH/Optum
Head of AI for TigerGraph (Silicon Valley)
Knowledge representation strategies for education
Using GenAI to create MicroSims and Intelligent Textbooks

The key to artificial intelligence is knowledge representation

Thinking About Personal Agents Since 2008¶

The Seed Email¶

Seed Email

From: Gang Qiu
Prof. Katie Zhao and I are organizing a monthly departmental workshop on AI for Research Productivity. Our goal is to build a community where graduate students can exchange ideas and explore how to use AI effectively, creatively, responsibly, and ethically in research.

Would you be available to give a 40-minute presentation at our upcoming workshop? ...

My Response¶

Title: From Researcher to Orchestrator: Managing a Team of AI Agents for Research¶

Abstract:¶

What if every researcher managed a team of 30 AI agents? In this hands-on talk I will describe a working model of research productivity in which each researcher coordinates 20 planning agents (which explore literature, decompose problems, and draft specifications) and 10 execution agents (which write code, run analyses, build simulations, and prepare figures and manuscripts). The researcher’s job shifts from doing every task to setting direction, supplying domain judgment, and reviewing results.

I will make this concrete with case studies from my own work, including interactive simulations and a full arXiv paper produced with AI agents, and a live demonstration of the two practical building blocks students asked about: plain Markdown files that capture a project’s goals and conventions, and reusable agent skills that turn a repeatable research task into a one-line command. We will also discuss how to keep this work responsible and ethical: verifying outputs, citing sources, and keeping the human accountable for the science.

The Problem: Research Funnel¶

The Solution: Researcher as Orchestrator¶

What Does Dan Use His Agents For?¶

Generation of interactive intelligent textbooks from a course description (10,000 hour task in 10 hours)
Generating simulations that live within these textbooks (MicroSims)
Generating reports, presentations and papers
Helping my mentorship program
Helping build my Coding Club hand-on kits (Smart Clock)
Promoting books, microsims and interactive infographic posters
Scheduling and calendaring

Using Agent Personas to Divide Tasks¶

Number One Question: Communication¶

How do you efficiency get all your Bots to communicate together as well as communicate with the agents in your department and at the University of Minnesota?

Recall: The key to artificial intelligence is knowledge representation

The Key: Personal Knowledge Graph¶

A central "ground truth" data structure to keep your content aligned with your goals.

Generating a Publishing Quality Paper¶

Ask your GenAI tool what the best practices are
- Length of paper
- Typical outline
- How many figures
- How many references
- Provide a template
- Create a build.sh script
- build.sh produces a main.pdf