Nguyen K. Phan: UHD Presentation

01Getting In

From HCC to UHD, to a Ph.D. at UH.

The application portal makes a Ph.D. look like a lottery. It isn't. Admission mostly comes down to one professor deciding they want you in their lab.

My UHD Experience. Small comfy classes, attentive professors, and a tutoring job at the Center for Math and Stats Support, my first taste of teaching. Hard work = high grades, no caveats!
Reach out early. Do not just send an application into the void. Find potential advisors whose work interests you and email them. I connected with my advisor, Dr. Guoning Chen, this way.
The 90% Rule. If you talk to an advisor, they like you, and they say they have a spot in their lab, you are 90% of the way there. They will provide the recommendation to the department.
The final 10%. Unless your grades or application are wildly out of bounds, having an internal advocate is the key to admission.
Ph.D. Requirements. The formal checklist: 30 credit hours of COSC courses, RCE Exams, Annual Reviews, and your Proposal and Dissertation defenses.

My Timeline & Stats

2014-2016: Houston Community College
2016-2018: UHD. Graduated Summa Cum Laude (3.77 GPA).
2018-2019: Researched smart gloves & RC drones for bridge integrity under UHD's Dr. Ting Zhang.
2019-2025: Ph.D. at UH. Completed core courses with 3.593 GPA.

02The Big Question

Do you want to pursue graduate studies?

Ask yourself:

Do you like working on problems that have no answer key? Not homework-hard, but genuinely unsolved. If that sounds fun rather than terrifying, a Ph.D. gives you years of time, funding, and an advisor to chase exactly those problems. If it sounds terrifying, that's useful to know too.

03Transferring to the University of Houston

Less paperwork than you'd fear.

The connection. Learned about UH Ph.D. programs through a friend and connected with Dr. Guoning Chen, now my advisor.
Research focus. Scientific Data Visualization and Modeling, specifically curve-based and mesh data.
Acceptance process. A few meetings where Dr. Chen probed what I knew and laid out what the lab expected. No trick questions.
Enrollment. Brief paperwork, transferred my I-20, and signed up for classes within a few weeks.

04Challenges at UH

A high hurdle from BS to Ph.D.

01 / The Hurdle

High Expectations

Jumping straight from a BS into a Ph.D. program is rough. Many students take a Master's first to bridge the gap. That's a valid path, not a detour.

02 / Coursework

Challenging Classes

Graduate courses assume you'll fill your own gaps. Nobody re-teaches the prerequisites. You catch up on your own time or fall behind.

03 / Publication

Submitting Papers

Your first submission will probably bounce. Mine did. The next section is about why that's normal, not fatal.

05The Reality of Research

Your effort is never wasted.

01 / Expect Rejection

It's an iterative process

Paper rejections are standard. Reviewer feedback is how the work gets better. You revise, you adapt, and you resubmit. A rejection is not a failure of your ability.

02 / Nothing is Lost

Shelved projects count

If a project stalls completely, you just shelve it. That work still goes into your thesis. No reasonable advisor will force you to stay an extra year just because of tough reviewers.

03 / Beyond the Code

Writing & Connections

Listen to your advisor! Expand your connections at conferences. Once you publish, make a working demo (HuggingFace, Colab) to get citations!

06Thesis Overview

The TL;DR

Towards Accessible and Robust Methods for Curve-Based Vector Field Visualization, Analysis and Synthesis

Six years of work, in four projects:

01 / Exploration

Curve Segment Neighborhood-based Vector Field Exploration via Graph-based representations.

02 / Generative AI

Physics-Constrained Vector Field Synthesis using Diffusion Models.

03 / LLM Integration

Human-in-the-loop LLM systems.

04 / Deployment

Flexible Cross-Platform Web Visualization (FCLWebVis).

The heavy math lives in a separate talk. Ask me about it after if you're curious.

07The Elephant in the Room

Is AI going to take every job?

Everything ships with AI in it now, whether you asked for it or not, and half the headlines say your job is next. Before you believe them, look at who profits from that story.

The Hype

Inflated Valuations

An AI company's valuation rides on what investors believe AI will do, not what it does. So the demo shows the best case, and the press release sells it as the everyday case.

The Scapegoat

Covering Financials

"AI can do it better" reads like progress. "Our revenue is down" reads like a sell signal. Some layoffs blamed on AI are the second sentence wearing the first.

The Parallel

The Dot-Com Bubble

In 1999 the internet was going to replace everything. Then the bubble popped, and the internet stayed useful anyway. It changed jobs; it didn't end them. Expect the same shape here.

The Takeaway

AI is a tool, not a boogeyman, and you can't judge whether a task is right for a tool you don't understand. If you like AI, learn how it works so you know when to trust it. If you hate it, learn how it works anyway: know your enemy. Keeping your head in the sand is the one option that gets you nothing.

08A Word of Caution

Don't rely too much on AI.

When using AI as an assistant, especially for writing or research, you must remain the expert.

Fact check, fact check, fact check.

You can ask AI to find relevant sources for a paper, but you have to check if those sources actually exist. Hallucinations are real. Use reliable, established sources to confirm anything an AI reports before you put your name on it.

09AI in Practice

Teaching complex graphics on an adjunct wage.

I taught three full-time courses. Concepts like ray-to-object intersection are incredibly dry on a PowerPoint slide. I wanted interactive, intuitive demos, but I didn't have the time to hand-code everything from scratch.

I used AI as an assistant to help build these web-based visualizations. It didn't get it right the first time: it almost never does. I had to iteratively refine the formulas and the output. But it turned abstract math into something students could interact with.

Two course sites. Every concept gets a demo you can poke at.

For COSC 2306 and 4370 I built dedicated lecture sites with live, interactive demos so abstract methods become something students can touch. Everything stays publicly accessible, anytime.

2306.pages.dev

COSC 2306 · Data Programming: lectures and interactive data-structure demos.

cosc4370.pages.dev

COSC 4370 · Computer Graphics: lectures and interactive rendering demos.

COSC 4370 · Interactive Computer Graphics

An LLM inside the 3D rendering pipeline

I designed an assignment where students control a live 3D scene through natural language. The real lesson is prompt engineering: students rewrite the system prompt to make the LLM manipulate the world more reliably. Interact with it during the talk!

3dstudio.pages.dev · COSC 4370 LLM exercise

CodeHelp · COSC 1336: guided programming help

CodeHelp guided AI help interface used in COSC 1336

COSC 1336 · Intro Programming

CodeHelp: an AI tutor that guides, never hands over answers

In my intro course I adopted CodeHelp, an AI assistant that scaffolds students toward their own solution rather than spitting out code. Students use it freely on exercises and practice exams; it lowers the anxiety of a first programming course without short-circuiting the learning.

Built by Somaia Alhazmi, a UH Ph.D. student researching AI in education. It's a live example of where a Ph.D. project can end up: in real classrooms.

10What's Next

From a UHD classroom to a career built on it.

Not long ago I was sitting where you are now. Last semester I taught three full courses at once and still built the demos you saw tonight, because I wanted my students to actually get it, not just survive the exam.

If you take one thing home: learn how AI works before you decide what it can't do, or what it will do to you. And if research pulls at you, email a professor whose work you find interesting. One email is how all of this started.

Thank You.