WingAI

Dating apps solved discovery and left conversation to fend for itself. The first message decides everything, and most people either send “hey” or send nothing at all. Matches expire, threads stall, and the apps have little incentive to fix it. That gap between matching and actually talking is what WingAI was built for.

Point WingAI at a screenshot of a match’s profile or an ongoing conversation and it writes openers and replies that sound like you, not like a bot. Its openers get roughly 2.5× the reply rate of typical first messages. Launched in 2023, it now runs on iOS and Android with a 4.8★ average across roughly 110,000 ratings.

Co-founder and the technical and product lead. I built the app and most of the systems around it:

• Product direction
• Mobile engineering (React Native / Expo)
• AI implementation
• Monetization & subscriptions
• Analytics
• Growth & creator/UGC systems
• User acquisition
• Operations

Over two million downloads, with ratings holding at 4.8★ as volume grew. That’s quality surviving scale, not just launch reviews. Distribution came from a creator program whose content has passed a billion organic views.

Scale forced the unglamorous work: keeping LLM costs sane at consumer volume, keeping generation fast enough to feel instant, and keeping output safe enough to stake a brand on.

Subscription-led, seven-figure annualized recurring revenue. Growth comes from the creator and UGC engine rather than paid-first acquisition: content that earns views instead of buying them, feeding a funnel we instrumented end to end.

The interesting problems were rarely the obvious ones:

• Iterating prompt quality against real reply-rate data, not vibes
• LLM cost and latency at consumer scale
• Subscription mechanics and pricing across two app stores
• Shipping fast with a very small team