MAGNI
Product design, Information architecture, AI integration, Prototyping · 2026

Quick / TL;DR
Problem
Two core systems had 17 structural issues and no clear next action
Assessments and Login & Onboarding end to end across 21+ IA nodes
It was a redesign of the entire enterprise ecosystem. I identified 2 core workflow gaps where AI, information architecture, design systems, UI and motion can meaningfully close the loop. It was identified by mapping: the dashboard, which gives every module equal visual weight; an assessment list that reads like an administrative table; and filters that behaved like navigation when they were really one page. Learners had no visible sequence through any of it.
Solution
Four decisions shaped the redesign. Three principles, rebuilt into four core flows.
Made key metrics visually dominant instead of competing for attention with everything else on screen
Reduced repeated metadata, showing ownership and status labels only when they matter
Led every dense screen with a visual summary before the table
Reframed assessments as a skill check with a visible sequence, not a form to get through

Three different people share one platform, learners trying to finish an exam, admins building and managing assessments, and customers accessing exam codes for their teams. One login screen served all three, and couldn’t tell them apart.
Learner
Three different people share one platform, learners trying to finish an exam, admins building and managing assessments, and customers accessing exam codes for their teams. One login screen served all three, and couldn't tell them apart.
Admin
Customer
Learning feels harder than it should
Hi, I am Kris, and I’m an Engineer



Five friction themes
Methods used: Heuristic audit, IA site mapping, and benchmarking against modern learning platforms
Platforms Selected
I looked at the platforms already running assessment heavy training, 360Learning, Docebo, Absorb, and Litmos, to understand what's working and where admins are still left to connect AI generated questions to a reusable library themselves.
What works
QGen generates quiz questions right inside the course editor, pulled from the content it's testing.
What's missing
Scoped to one course at a time. No path from a generated question into a library other admins can reuse.
What works
Docebo Shape turns external documents into training content automatically, with role based access built in.
What's missing
Generates new material, but the output doesn't feed into a shared question library.
What works
A centralized Question Bank lets admins reuse questions across assessments, and Absorb Create AI can generate questions from content.
What's missing
To be written
What works
Built in AI quiz generation with a custom login page and defined roles for employees, customers, and partners.
What's missing
Separate audiences are supported at setup, but reviewers note keeping each role's access current takes ongoing admin work.
Hover to reveal
40% to 97%
Login usability score
Outcome
01
Hover to reveal
6 to 8* mins
Assessment generation completion time
Outcome
02
Structure is the deliverable
This project pushed me to see structure as the actual deliverable, not the visual layer on top. The hardest part wasn’t redesigning the screens. It was deciding where ownership, status, and sequence actually belonged, and stripping out everything that didn’t earn its place on screen.
Enterprise work is its own discipline
Working inside an existing enterprise product is its own discipline. Real users, and constraints, and a team that has to live with every decision after I leave. A lot of the redesign came from sitting with the friction long enough to see the pattern underneath it. That’s both the harder path and the one that actually holds up.
I sculpt motion out of nothing.
- 3D modeling
- Particle simulation
- Lighting
Still here?