Large language model (LLM) enhanced digital representation of smell for food manufacturing
Supervisors: Tanoj Singh (CSIRO), Chen Wang (CSIRO), Xavier Sirault (CSIRO), and Rod Peakall (ANU).
Recent research has shown natural language processing techniques and graph neural networks can be used to learn a digital representation of smell from odorous molecules. Large language models (LLMs) like GPT/Llama provides further opportunities to make smell representation suitable for practical use in food manufacturing. These machine learning techniques have potential to change food manufacturing processes by paving the way for optimized seed selection and protein fractionation to achieve desired food product properties.
The PhD candidate will develop novel machine learning based methods for smell representations of molecules of interest. The candidate will then use these representations to develop an understanding of the biological foundation of odour perception and develop new methods to produce a smell map.
The generic digital approach developed for the representation of volatile organic compounds will be applicable to food manufacturing and to wider biological research such as pollination. The selected candidate will be carrying out research work both at CSIRO (laboratories at Black Mountain, Canberra and Melbourne) and ANU.