LaunchJury
beaconA command-line agent that reads a repository, drafts a complete product launch and tests the drafts against a jury of seven audience personas before you post.
A real-time voice AI for banking customer service that handles interruptions and detects emotion; the team reports responses in under 300 ms.
Now a days spending on human telecallers is expensive as well as inefficient, humans can make mistakes as well. However making a voice bot that actually works real time, understands emotions and answers accordingly is something which was missing. Traditional voice AI systems can feel slow and do not handle interruptions well. This is especially a problem for banking customers using voice-based customer service, where users expect quick and natural conversations.
miniVoxSetu is a real-time voice AI platform that listens to the user, converts speech to text, understands the query using an LLM and RAG and gives a voice response. It uses streaming, fast retrieval, acoustic analysis, and a 7-layer interruption system to reduce delays and quickly stop the AI when the user speaks. This brings the response time from around 1.2 seconds to under 300ms, making conversations faster and more natural. It is also has BargeIn functionality and emotion detection using librosa and HuBert.
Not documented in the submission.
BugBiceps
Individual builders are not listed yet — names are published only with their permission.