Krystel Reina

Product designer
for AI.

I design ways for people to understand complex systems, question an AI’s reading, and decide what happens next.

Enterprise practice

The work behind
the experiments.

I work in enterprise UX and systems design at a Fortune 100 agricultural cooperative, with earlier product work in healthcare SaaS. That work brings people, platforms, and different understandings of the same problem into the same room.

The independent projects below let me examine particular questions: how a person corrects an AI reading, how advice stays separate from a decision, and how a complex field becomes easier to learn.

Read my professional background

Selected work

Human judgment, made tangible

Start with a correction, then a decision. These public examples show my product direction and the design questions I pursue with AI assistance.

01 · ThoughtGlass

Making an AI interpretation
correctable.

When a system’s reading misses the point, a person needs a way to change the relationship it has drawn.

My contribution: I directed the product’s development, required traceable states, and asked for motion to explain relationships. When the correction loop froze during use, I called for an audit.

Independent project · AI-assisted · Worked interaction demo

Inspect how a proposed reading becomes a visible correction, with the earlier interpretation retained.

Read the case study Evidence and limits
ThoughtGlass’s corrected graph connects system stability, a component contract, engineering, and designers; the rejected reading remains above it.
The supplied engineering example, after correction.
Actual demo capture · September 8, 2026

02 · AI Council

Advice stays visible.
The decision stays separate.

When specialists disagree, a single polished answer can hide the very difference a decision-maker needs to inspect.

My contribution: I specified reusable roles, visible disagreement, decision history, and human approval points. Codex contributed substantial technical design and implementation.

Independent project · Local prototype · Synthetic worked example

Follow the worked example

Making a complex field learnable

03 · The AI Design Atlas

The Atlas Explorer arranges creative AI tools into families, with a role, strength, and failure mode on each field card.
Atlas Explorer · Actual interface capture, September 8, 2026. Tool descriptions are the Atlas’s own framing.

The AI Design Atlas

A way into a complicated field.

A visual language for understanding creative AI tools and imagining how they might work together. Read an element, combine it with others, then examine the idea through different artifacts.

My contribution: I defined the audience and learning purpose, chose reusable theme foundations, and identified where the interface failed to carry those themes through.

Independent project · Public explorer

Explore the Atlas

Further explorations

UX RAVE

Research context turned into structured artifacts, with prompt instructions against invented evidence. Generated content still needs review.

Working prototype

StarGlass

A depth-psychological astrology interpreter, explored as both a portable skill and a web experience.

Private preview

The Theme Studio

A motion and theme engine inside the Atlas. An exploration of changing expression while keeping meaning recognizable.

Open the Atlas, then choose Studio

Inside the Atlas

Field notes · Essays · Receipts

The thinking the instruments run on.

Private language, context packs, and one idea run through a music model eight times. The essays and their prompts, transcripts, and audio receipts live here, together.

Start a conversation

What are you trying
to make understandable?

Working through a complex workflow, an AI interaction, or a design question that crosses teams? I’m interested in making the next step clearer together.