A networking experience that breaks through transactional exchanges by transforming how people share themselves - designed, prototyped, and built to make meeting new people memorable instead of forgettable.
Timeline
My role
Product Designer & Developer
Tools
Figma, React, TypeScript, Lovable, LLM (Gemini + OpenAI)
Career fairs. Networking events. Industry mixers. They all follow the same script.
You hand out resumes, promise to connect on LinkedIn, exchange the same intro over and over. And then you leave exhausted, with a vague memory of who you talked to.
A few days later, you're scrolling through LinkedIn trying to remember: was that the person from the fintech company or the consulting firm? Did I already send them a request? What did we even talk about?
Meanwhile, the people you met saw 50+ others that day. You're just another LinkedIn notification.
I grew up collecting Pokemon cards. Trading with friends felt exciting - you wanted to see what they got, you showed off yours, you remembered people by their cards.
What if professional networking worked the same way? If introducing yourself felt like showing off a collectible card instead of handing over a resume, people would actually remember you.
I designed and built a networking experience that turns introductions into collectible trading cards. Built end-to-end in React + TypeScript with machine learning pipelines for personalized content generation.
I knew this could work in theory. So I tested it in the real world. I brought Carded to a networking event.
What I observed:
Ice broke faster with cards than traditional introductions - conversations started playful, then naturally became substantive
People remembered me - the format made me stand out and follow-ups happened organically
But some friction points emerged: people didn't immediately understand the QR scan flow, some AI-generated facts were too generic, and occasionally the AI generation would fail or timeout
What I changed after:
Made QR code more prominent with visual cue ("Scan to collect")
Refined AI prompts for more personality-driven, conversation-starting facts
Built dual AI pipelines with graceful fallbacks - if one model fails or times out, the system switches to the backup to ensure cards always generate
Retro Gaming Aesthetic as an Icebreaker
After some trial & error, I landed on neon green color palette, pixel art and accent decorations that reference classic arcade games and evoke nostalgia. The aesthetic had to feel playful but intentional, not gimmicky
Personalized Fun Facts That Break the Ice
Here's the concept: The answers to the quiz generate exaggerated, obviously-fake-but-personality-revealing facts. They're based on what the user shares, but amplified to be funny.
Technical Implementation: Built dual ML pipelines (Gemini 2.5 Flash + GPT-4) that analyze quiz responses and generate personality-specific content. Tested multiple models to find which created the most engaging, shareable results. Built graceful fallbacks in case either pipeline fails.
Questions Designed for Low Friction
The 5-question experience isn't just data collection - it's the first impression of Carded. I spent as much time crafting the copy as I did on visual design.
Currently Exploring: Physical Smart Cards
The vision: E-ink "smart cards" (~3-4" display) that show your card with NFC tap-to-share. Update wirelessly, lasts weeks on battery. Bridges digital convenience with physical presence at events.
Additional Improvements Planned:
Onboarding flow: First-time users need clearer guidance on how to use cards effectively at events
Collection organization: Users want to categorize cards by event or add personal notes
Follow-up prompts: Nudge users to actually reach out to connections they made
Accessibility: Ensure card designs work for users with visual impairments
