subZeroSearch

Inspiration

The idea for this project came from the difficulties recruiters face when sorting through many resumes to find the best candidates. We aimed to use AI to make this process quicker and more accurate, helping to improve hiring decisions by offering deeper insights into candidate profiles.

What it does

SubZeroSearch is an AI-driven recruitment tool where recruiters can upload resumes and get detailed analysis on candidates, including an AI chatbot for interaction. It pulls out key details like skills, experience, and accomplishments, presenting them in a clear way. It also suggests suitable job roles for candidates based on AI analysis.

How we built it

We developed SubZeroSearch using Streamlit for the interface, Snowflake for data management and retrieval, and Cortex LLM's for the chatbot functionality. The platform uses the 'Mistral-large2' model for both Cortex and Snowflake's search capabilities.

Challenges we ran into

Some of the challenges with Snowflake's Cortex service were related to quickly indexing new files, which takes at least 60 seconds. This was temporarily solved with a loading UI in Streamlit to wait for users for a minimum of 60 seconds before starting their chatbot. Additionally, it was initially difficult to get structured responses, but combining the 'Mistral-large2' model with Snowflake's Snowpark complete function made this task much easier.

Accomplishments that we're proud of

We are proud to have built a platform that makes recruitment easier and provides useful insights to recruiters.

What we learned

We learned a lot about integrating AI, handling data, and designing user experiences. We understood better how to manage unstructured data and optimize app performance. We also learned the value of user feedback in enhancing our tool.

What's next for SubZeroSearch

Moving forward, we plan to upgrade SubZeroSearch with more sophisticated AI models for better insights. We also want to add more comprehensive analytics and reporting features to the platform.

Built With

• cortex
• mistral
• snowflake
• streamlit