Year
2025 - Present
Company
korpus
Role
Builder
Login & Onboarding Flow
Parents had trouble signing up, especially from shared links. We’re testing a simpler flow where kids invite parents through a one-tap Face ID link.
Feedback in Action
Charlotte mentioned Duolingo’s family plan — parents want to feel included and part of their child’s membership.
Email Feedback Example
Users can share feedback right from the app by email,(hello@korpus.app) helping us improve quickly and focus on what matters most.
Logo concept (eye-in-frame): scanning + protection.
Color: Trust blue, with state colors (green safe, red scam, orange suspicious).
Type: Clean, rounded geometry for readability and warmth.
The logo is built around the metaphor of an eye within a frame — a symbol for scanning, detecting, and keeping watch.
The eye represents vigilance and safety, echoing the product’s purpose of protecting users from threats.
The brackets / scan frame suggest the act of focusing in, just like a camera or OCR tool would when analyzing text.
This duality communicates the core function: Korpus looks closer so users don’t have to.
The wordmark uses a clean, geometric typeface that feels both modern and trustworthy. Rounded edges and balanced proportions soften the tech-heavy nature of AI, making the product feel human-first.
In early testing, users expressed concern about overly “technical” brands — so I opted for a system that feels protective without being intimidating.
Color & Visual Language
• Blue was chosen as the primary color for its strong association with trust, security, and reliability.
• Accent colors (green, red, orange) extend into the product’s scan states, reinforcing clarity: safe, dangerous, or suspicious.
• The identity system flexes from friendly and minimal (light backgrounds, clean typography) to strong and alert (black backgrounds, bold contrasts) depending on context.
Design rationale bullets (concise)
Color-coded states reduce cognitive load
Confidence % adds transparency
Red flags / positive indicators teach patterns
Next-step guidance closes the loop
Design Rationale
Color-coded feedback states (green, red, orange) makes risk level instantly recognizable, reducing cognitive load in stressful moments.
Confidence percentage adds transparency to the AI’s decision-making.
Contextual explanations (red flags vs positive indicators) help users understand why the result was given.
Action-oriented guidance ensures the experience doesn’t stop at detection but empowers users to take protective steps.
Korpus uses a combination of AI, OCR, and pattern recognition trained on real-world scam messages. It analyzes:
Language tone and urgency (phishing patterns)
Suspicious links or spoofed domains
Repetitive scam phrases flagged in previous reports
Our proprietary system continuously learns from user reports. The more people scan, the smarter Korpus becomes — creating a collective shield of awareness.
The home page gives users an immediate sense of control — showing their scan history, safe vs scam breakdowns, and even global scam trends.
For many testers, seeing how many others faced similar scams was reassuring: it turned isolation into community.
Email phishing: domain mismatch, link/attachment risk, impersonation cues
SMS/text: unknown sender, urgent tone, shortened links
Voicemail/robocall: automated voice, spoofed IDs, payment demands
Consistent structure, familiar words, and “one clear action” per state reduce panic and help people move forward.
Trust & Safety Decisions
Trust Principles
Human-first tone: calm, direct, non-technical
Explainability: show reasons, not just verdicts
Community value: every scan improves the model
Optional feedback: “See a mistake?” for corrections
Privacy: scans are anonymized; no contacts harvested
To simplify results, I designed three clear states — Safe, Suspicious, and Confirmed Scam — each color-coded and paired with confidence levels and next steps.
The results page is also the emotional center of the product.
It’s where users get answers — Safe, Suspicious, or Confirmed Scam — each paired with a confidence score, clear explanation, and next steps.
Color-coded feedback (green, orange, red) and plain-language summaries turn technical AI output into emotional clarity.
Design goals:
Transparency (show confidence levels)
Education (why it was flagged)
Actionability (what to do next)
Within the first month of beta, Korpus saw:
Dozens of heartfelt emails from users sharing how Korpus helped protect them or their parents
Role & Scope
Product strategy, flows, interaction design
UX writing & content standards
Brand identity (logo, color, typography)
MVP research & early tester program
user interviews
everything
Constraints
Small, fast beta (limited engineering)
Must work with screenshots (OCR) + copy/paste
Audience skew: older adults + busy people
Role & Scope
Auto-scan from share sheet (one-tap from Messages/Mail)
Breach & brand alerts (if an org you use is actively spoofed)
Family mode (share results with a trusted contact)
Deeper community signals (trending scams in your area)
Designing for high-stress moments is about reducing decisions and increasing clarity.
Transparent language, consistent patterns, and one obvious next step build trust faster than any “smart” UI.
