Solvearn needed rapid distribution and user growth, but consistent content output was a bottleneck. Manually finding viral formats, extracting patterns, and turning them into scripts was too slow to post at the volume required to win attention. They also needed smarter user-to-user matching so the product experience actually retained the audience they were driving in.

We built two systems that worked together: a short-form content research and scripting engine that continuously identified viral Instagram reels, extracted winning patterns, and generated ready-to-record scripts, and a recommendation pipeline using vector search to improve AI matching inside the platform, so new users had a better "aha" moment after they arrived.