Inspiration

Chef Gusteau grew out of my own frustration with not knowing how to cook even simple meals. I always dreamed of having a real chef beside me in the kitchen—someone I could ask questions, learn from, and turn to when something went wrong. Chef Gusteau is my attempt to bring that experience to life.

What it does

For the Nebius x NVIDIA Global AI Hackathon, we chose the Personal AI track and built Chef Gusteau—a voice-enabled cooking assistant that takes you from “What should I make?” to dinner on the table.

Talk to Gusteau to choose recipes, substitute ingredients, and get step-by-step guidance, supported by an interactive visual interface.

It combines self-hosted Nemotron speech recognition and text-to-speech with a Nemotron LLM running on Nebius Token Factory.

How we built it

We used React, TypeScript, and Tailwind CSS for the interface, with a Python server deployed on a Raspberry Pi.

The voice pipeline connects Nemotron speech recognition, Nemotron Super 120B through Nebius Token Factory, and self-hosted Magpie text-to-speech. Application state tracks recipes, substitutions, completed steps, and timers.

Challenges we ran into

Maintaining conversational context was a major challenge: “swap it with Swiss” needs to reference the ingredient discussed earlier. We also worked through transcription errors, browser audio permissions, and connectivity between the Pi and model services.

Accomplishments that we're proud of

We connected recipe discovery, conversational customization, and guided cooking in one working application. Gusteau updates recipes, remembers confirmed steps, controls timers by voice, and surfaces helpful images and videos—all through a Raspberry Pi deployment.

What we learned

A useful assistant needs reliable state and recovery mechanisms alongside a capable model. Remembering context, handling misunderstandings, and combining voice with clear visual feedback are essential to earning users’ trust.

What's next for Chef Gusteau

We want faster conversations, better handling of ambiguous requests, more recipes, and deeper personalization. We also plan to coordinate multiple dishes so Gusteau can help users prepare an entire meal.