A voice-activated smart home control system integrating the Google Gemini API for natural-language command parsing, driving Arduino-controlled devices directly. Ranked in the top 12 of more than 250 teams at GDSC DevJams 2024.
Voice-activated home controller that interprets natural-language commands and drives devices directly, no fixed grammar.
An LLM instead of a command grammar
The Gemini API parses natural-language commands, so the system resolves intent rather than matching phrases. That removes the memorised vocabulary that makes most home automation feel brittle — a request works because it means something, not because it was worded correctly.
Straight through to the hardware
Parsed intent drives Arduino-controlled devices directly. Keeping the path from language to actuation short is what stops the round trip feeling like a query rather than a control action.
Media as a first-class target
The Spotify API is wired in alongside the physical devices so that playback is one of the things a command can address, rather than a separate app the same voice request cannot reach.
Built under hackathon constraints
The whole system was built and demonstrated at GDSC DevJams 2024 and placed in the top 12 of more than 250 teams — the scope was set by what could be made to work end to end in the time available.
Open-language commands
No trigger vocabulary to memorise.
Direct device control
Intent through to Arduino actuation.
Media control
Spotify addressable by the same voice path.
Competition-proven
Demonstrated end to end under hackathon conditions.
Parsing: Gemini API
Devices: Arduino
Result: Top 12 / 250+
Divyakush Punjabi