01 / 03
TRACENERVE
See what AI agents do during a run, before a deviation turns into rework.
- Role
- Senior product designer, led end to end
- Team
- UX/UI designer, PM, 2 frontend, 2 backend, AI/ML, DevOps, QA
aleh pankou
Seniorproductdesignerfor AI
Product designerAI interfacesDesign systemsResearch to handoff
About me
I’m Aleh Pankou, a senior product designer based in Chicago, working remotely with a team in San Francisco. For six years I have designed one product from the inside at PASV, a coding academy that trains IT specialists and builds its own software and AI tools.
I grew there from junior to senior and have owned the AI layer of the platform since early 2025. I design for outcomes, not screens. Time is a company’s most valuable resource, so I bring a working prototype to engineering before anything is built, and together we make sure the product meets every requirement the first time: results for the business, and what people were waiting for.
- 6
- years in product design
- 120
- components in one shared Figma library
- 300+
- screens audited for WCAG 2.1 AA
- +12%
- course purchases after a payment-screen redesign
Scroll downto begin the story
[ Experience ]
Experience guides
the work
PASV · 2020 — now
[01]
Senior Product Designer
Aug 2024 — now
Sets the standard for 12 designers, PMs and marketers: weekly reviews, the style guide and voice rules. Owns the AI layer of the platform since early 2025.
[02]
Product Designer
Mar 2022 — Jul 2024
Took whole products from research to handoff and worked side by side with engineering on the system they are built from.
[03]
Junior Product Designer
Sep 2020 — Feb 2022
Learned the product by designing it end to end: user research, analysis, prototypes and documentation in Figma.
[ Selected work · 2026 ]
Two AI products and one healthcare product, designed end to end.
- 01TRACENERVEAI agent observability
- 02PATCHBOUNDAI agent work verification
- 03WayFindDoctor booking, price first
02 / 03
PATCHBOUND
Verify AI-agent work and find errors, unintended changes and regressions.
- Role
- Senior product designer, led end to end
- Team
- UX/UI designer, PM, 2 frontend / canvas, 2 backend, 2 AI/ML, DevOps, QA automation
03 / 03
WayFind
Book a doctor in the US and see the real price before the book button.
- Role
- Product designer
- Team
- UX/UI designer, team lead, 2 frontend, 2 backend, QA
[ What I do ]
Design that makes
AI clear
Focus
[01]
AI interfaces
Dashboards and flows that show what an AI agent did, what it needs, and where a person decides.
[02]
Design systems
A 120-component Figma library built with a front-end engineer, so design and code use the same parts.
[03]
Research to handoff
Interviews, flows, prototypes and specs, carried end to end inside one product team.
[04]
Accessibility & QA
300+ screens audited against WCAG 2.1 AA, with the critical issues fixed before release.
Portfolio 2026
Based in Chicago, ILOpen to new roles
Scroll
Chapter I · About
I’m Aleh Pankou, a product designer with six years in product teams. I live in Chicago, work remotely with a team in San Francisco and design end to end in Figma: research, flows, prototypes, design systems and documentation.
At PASV, a coding academy that builds its own software and AI products, I grew from junior to senior product designer. Today I design how AI agents show their work to learners, and I write the prompts and system instructions behind them.
People
product &
AI agents
PASVFigmaClaude CodeWCAG 2.1 AAIT-Academy MinskClaude Academy
Design
Explain
AI Build
What I doProducts · 6 yearsDesign systemsSince 2025
How I work
I design for product teams that ship AI features and want people to understand them.
- Work
- Authorized in the US, no sponsorship needed
- Relocation
- Open to relocation
Chapter II · Screens & experience
Screens that tell the story
9Selected screensThree products · 2026
Clearwhen itmatters
Quietwhen itworks
Readable,honest &shipped
Designedto be read
built
to be trusted
built
to be trusted
Every screen has to earn its place: clear about what the AI did, honest about what it does not know, and quick to hand the decision back to a person.
PASV · Programming Academy in Silicon Valley · Sep 2020 — now
From junior to senior, inside one product.
A coding academy that trains IT specialists and builds its own software and AI products. Six years, three titles, products designed end to end.
- Sep 2020 — Feb 2022Junior Product Designer
Learned the product by designing it end to end: user research, analysis, prototypes and documentation in Figma.
- Mar 2022 — Jul 2024Product Designer
Took whole products from research to handoff, and worked side by side with engineering on the system they are built from.
- Aug 2024 — nowSenior Product Designer
Set the standard for 12 designers, PMs and marketers: weekly reviews, the style guide and voice rules. Since early 2025, the AI layer of the platform.
Inside PASV · what changed after my work shipped
One library the whole product stands on.
I built and maintained a 120-component Figma library together with a front-end engineer, so design and code use the same parts. Then I audited 300+ screens against WCAG 2.1 AA and fixed the critical issues.
120
components in the library300+
screens audited for WCAG 2.1 AA~15
critical accessibility issues fixedFewer surprises between design and code.
I introduced a handoff checklist that engineers run before a build goes to QA. I redesigned the launch approval chain: who signs off, and in what order.
~15→2–3
critical QA returns a month4–5→2
days to approve a launchA product that explains itself.
I wrote 45 help-center articles where learners got stuck, and rewrote the onboarding email sequence. I wrote the content style guide, voice rules and glossary for 12 designers, PMs and marketers, and ran weekly reviews for the same 12 people.
−34%
“how does this work” tickets in the first quarter45%→62%
two-week activation12
people writing in one voiceTeaching the product how to show AI.
I wrote prompts and system instructions for the platform’s AI agents, and evaluated and rewrote their output every day. I wrote guidelines for 9 patterns that govern how AI output is shown to learners.
The hard call: a two-week test across 4,800 sessions showed that per-step confidence scores raised repeat failures. I replaced the scores with plain-language step states.
23→6
one-off AI patterns a quarter22%→28%
repeat failures while scores were shown, so they wentThree titles, one habit: make the product explain itself.
Designthatexplains.
2024 — now
Senior Product Designer, PASV
2022 — 2024
Product Designer, PASV
2020 — 2022
Junior Product Designer, PASV
6Years
in product
design
in product
design
120Components
in the design
system I built
in the design
system I built
300+Screens
audited for
WCAG 2.1 AA
audited for
WCAG 2.1 AA
62%Activation
after onboarding
redesign, from 45%
after onboarding
redesign, from 45%
Results at PASV · 2020 — now
Six yearsinside onegrowingproduct.
What changed after my work shipped. Before on the left, after on the right.
- 01~152–3a monthHandoff checklistA checklist engineers use before a build goes to QA.Critical QA returns
- 024–52daysApproval chainRedesigned who signs off and in what order.Launch approvals
- 03−34%in one quarter45 help-center articlesAnswers written where people get stuck.“How does this work” tickets
- 0445%62%of new usersOnboarding emailsRewrote the onboarding email sequence.Two-week activation
- 05236a quarter9 AI output patternsGuidelines for how AI output is shown to learners.One-off AI patterns at review
- 0622%28%so I removed the scoresPlain-language step statesA 4,800-session test showed confidence scores hurt, so I replaced them.Repeat failures when scores were shown
- 07~15fixed across 300+ screensWCAG 2.1 AA auditAudited 300+ product screens built on the 120-component library.Critical accessibility issues
Selected work · 202601 / 03
A view into what AI agents actually do.
Problem
A trace answers none of the questions people ask.
Agents drifted from the brief, and the team found out only after the run, when fixing it meant rework or a rollback.
57%of surveyed teams run agents in production (LangChain, n=1,340)
Research
Started inside the company.
About 30 interviews inside and outside PASV, surveys inside the company, 6 observability tools reviewed, domain research on agent telemetry.
Insight → Decision
A trace is raw material, not an answer.
So a run reads as situation, what each agent did and the evidence, and the raw trace stays one click away.
Outcome
A run you can read.
A working prototype inside PASV. Team feedback: tasks done about 3× faster, with fewer reruns, because drift is caught during the run.
- Role
- Senior product designer, led end to end
- Platform
- Desktop web · internal
- Year
- 2026
Emailalehpankou0110@gmail.com
LinkedInlinkedin.com/in/aleh-pankou
Based inChicago, IL · remote for San Francisco · open to relocation
WorkAuthorized to work in the US · no sponsorship needed
Looking forSenior Product Designer · AI Product Designer · UX/UI