This project was inspired by a simple observation: people constantly express what they need online, but brands rarely hear them. Posts like “my feet are killing me” or “my phone just died” are full of buying intent, yet traditional social listening tools can’t detect them. We wanted to change that — to build an AI system that listens not for keywords, but for meaning.
We built a real-time matching engine that ingests social posts, vectorizes their text, and compares them against a database of intent phrases generated by a large language model. When a match is found, an AI sales agent crafts a contextual, human-sounding reply that connects the user to the right product or service.
We learned how powerful vector search and LLMs can be when combined — and how critical fine-tuning thresholds and phrasing are to keep interactions natural, not robotic. The biggest challenges were optimizing for speed (processing thousands of posts per second) and ensuring AI responses felt authentic.
In the end, we built something that feels like the future of marketing: a system that transforms social conversation into real-time commerce by understanding intent, not just words.
Log in or sign up for Devpost to join the conversation.