Description
Voice-controlled smart home systems are becoming increasingly intelligent
thanks to advances in AI. However, many commercial solutions rely on cloud
processing, raising privacy concerns due to the transmission of sensitive voice
data. The goal of this internship is to develop a fully offline voice assistant for
smart home control. An MCU-based satellite device will capture voice commands
and perform wake-word detection, while a local server will handle speech
recognition and intent understanding using AI models. The resulting intents will be
used to control smart home devices such as a thermostat. The focus is on privacy,
low latency, and seamless user interaction.

Your task
• Select and evaluate AI models for offline speech recognition and intent
detection.
• Build and configure the AI software stack on a local server.
• Implement wake-word detection and audio transfer from the MCU satellite.
• Integrate smart home device control based on the detected user intents.

Contact
For more information, please contact Tom Lierman: [email protected]