Edoardo Conte Transforms Hospitality Recruitment at Restworld with Real-Time AI
Edoardo Conte, Co-Founder and CTO at Restworld
Co-Founder and CTO at Restworld
Edoardo Conte is Co-Founder and CTO at Restworld. He is responsible for the software development at 360 degrees, from team management, to web app development to machine learning. We currently work on a user-facing web application and an internal management system, which helps us in finding the right candidates for a given job position.
Transcript
Restworld supports employers and employees in the records sector in finding each other. Hotels, restaurants and bars make the records market. So we are vertical in that industry domain and support them to find each other. So we are a mix between a job board and a recruitment agency. We offer a competitive advantage to our users because they will have a completely different experience in their segmentation and preference setting, so they can express exactly who they are and what they are looking for. We try to give them this kind of experience.
Right now we are more in the generative AI. We've just started our journey in that direction. So that we create embeddings, we heavily use embeddings and vector search to give some semantic interpretation of our data that allows us to create this knowledge representation of it. And we could traverse all the data to extract all the needed information that can be used by external tools like LLMs or other pieces of technology like orchestrators and so on.
In our strategy, we wanted to implement Gen AI and these kinds of technologies, and we found in DataStax a partner that could support us in providing a product and guiding us through the building process. So what we actually implemented was to embed the job offers that we have in our system. Imagine ourselves as a marketplace. So on one side, we have job offers and job positions. On the other side we have workers. We concentrated only on job offers since we have much more control over them. We also used text embedding models to create vectors out of them. And we stored those vectors on Astra DB. Now on top of those vectors, we can navigate through them. We can search through them.
We implemented a matching algorithm, which is a sort of collaborative filtering algorithm, in which by giving a job offer, we could find the most similar job offer that we've managed in the past and associated with them, we can fetch the workers that were shortlisted or hired for those job position, and we assume that those workers, or at least some of them, may be interested in the job offer that we have opened right now. So we create this sort of recommendation algorithm.
Restworld supports employers and employees in the records sector in finding each other. Hotels, restaurants and bars make the records market. So we are vertical in that industry domain and support them to find each other. So we are a mix between a job board and a recruitment agency. We offer a competitive advantage to our users because they will have a completely different experience in their segmentation and preference setting, so they can express exactly who they are and what they are looking for. We try to give them this kind of experience.
Right now we are more in the generative AI. We've just started our journey in that direction. So that we create embeddings, we heavily use embeddings and vector search to give some semantic interpretation of our data that allows us to create this knowledge representation of it. And we could traverse all the data to extract all the needed information that can be used by external tools like LLMs or other pieces of technology like orchestrators and so on.
In our strategy, we wanted to implement Gen AI and these kinds of technologies, and we found in DataStax a partner that could support us in providing a product and guiding us through the building process. So what we actually implemented was to embed the job offers that we have in our system. Imagine ourselves as a marketplace. So on one side, we have job offers and job positions. On the other side we have workers. We concentrated only on job offers since we have much more control over them. We also used text embedding models to create vectors out of them. And we stored those vectors on Astra DB. Now on top of those vectors, we can navigate through them. We can search through them.
We implemented a matching algorithm, which is a sort of collaborative filtering algorithm, in which by giving a job offer, we could find the most similar job offer that we've managed in the past and associated with them, we can fetch the workers that were shortlisted or hired for those job position, and we assume that those workers, or at least some of them, may be interested in the job offer that we have opened right now. So we create this sort of recommendation algorithm.