Voice Task Assistant

Voice Task Assistant is a voice-based AI agent built for the AssemblyAI Voice Agent Hackathon. It demonstrates a complete pipeline from raw audio input to a real, saved action. The project works in three stages. First, it takes an audio recording and sends it to AssemblyAI's speech-to-text API, which returns an accurate transcription of what was said. Second, the transcribed text is analyzed using intent detection logic to determine what the user wants — specifically, whether it's a reminder request. The system is built to recognize several natural phrasings, such as "remind me to," "remind me about," and "reminder to," rather than relying on one rigid exact phrase. Third, once a reminder intent is detected, the relevant task is extracted from the sentence and saved to a local JSON file along with a timestamp, with a confirmation message shown to the user. This project was built solo, from scratch, by a third-year AI and Robotics Engineering student with no prior hands-on coding experience before this hackathon. It represents a genuine, functional use of AssemblyAI's transcription technology as the core of a working voice agent, and reflects real problem-solving along the way, including securing API keys using environment variables and handling various natural language phrasings for reminders. Future improvements could include support for additional intents beyond reminders, real-time voice input instead of pre-recorded files, and integration with actual notification or calendar systems.