What it does
TalentWave.AI is an innovative application powered by machine learning that revolutionizes the way companies handle incoming job applications. It streamlines the resume screening process by efficiently categorizing and assessing resumes submitted for job openings. It allows employers to rank candidates based on specific skills they are looking for, which makes it easier to filter out less qualified candidates, thus reducing the amount of CVs that the employer needs to look through, as they can receive hundreds to thousands.
How we built it
Our solution was designed using a combination of technologies. On the backend, we used Python, leveraging the sentence_transformers library integrated with PyTorch. We took the extra step of fine-tuning the all-mpnet-base-v2 model using a comprehensive dataset of resumes. Additionally, we employed the pdfplumber library to seamlessly extract text from PDF resumes. The data, along with job descriptions, was stored in an SQL database, ensuring accessibility and organization. On the frontend, we used Angular.js, coupled with the Material UI framework, to create a user-friendly interface for TalentWave.AI. This dynamic front-end interface seamlessly interacts with the backend, providing real-time updates for each job posting created. The use of RX.js allowed us to use observables, enabling our application to dynamically respond to changes in job postings. This means that as new job postings are created or updated, the interface automatically updates in real-time, providing an efficient and intuitive user experience. We also incorporated Flask, a micro web framework for Python. Flask served as the bridge that seamlessly connected our machine learning backend with the frontend. It allowed us to create a responsive and efficient web application that can handle incoming requests, process data, and deliver results.
Challenges we ran into
While developing TalentWave.AI, we encountered a significant challenge during the fine-tuning process. Initially, the model was producing many NaN (Not-a-Number) results, which confused us. It required several hours of dedicated effort to find the issue, but we ultimately resolved the problem with persistence.
Accomplishments that we're proud of
Our team is immensely proud of the tangible results we achieved with TalentWave.AI. For many of us, this project was the most we've ever delved into machine learning. For others, it was an opportunity to delve into frontend development. Either way, we all learned a lot, and actually created something that works quite well.
Log in or sign up for Devpost to join the conversation.