Conversational Text-to-SQL Agent

At Innowhyte, we receive a large volume of candidate resumes for each job role through our recruitment pipeline. While only one candidate may be selected for the job, this doesn't imply that the others were entirely unsuitable. Rejections can happen for various reasons—such as salary expectations being too high, or another candidate being a better fit for a specific client project.

Since we retain all candidate profiles in our database, we wanted to leverage this existing pool before sourcing new applicants for similar roles in the future. To do this effectively, we needed a system that could search, compare, and filter candidates based on contextual criteria.

Additionally, we aimed to give stakeholders the ability to quickly access insights—such as the number of candidates added to the pipeline, how many were rejected, and other key statistics—along with the option to compare candidates and schedule interviews seamlessly.

Current Problems

• Inefficient Reuse of Existing Candidates:
  Although many candidate profiles remain in the database after initial rejection, there was no streamlined way to efficiently rediscover or match them to new openings—leading to duplicated sourcing efforts and slower hiring cycles.
• Inefficient Stakeholder Access to Data:
  Business users and stakeholders struggled to retrieve statistics like candidate pipeline volume, rejection counts, or hiring trends without technical support.

Solution

We built a conversational AI system that supports the following:

• Enables recruiters and stakeholders to query the candidate database using plain English, removing the need for complex filters or SQL knowledge.
• Allows users to search for candidates contextually, such as "show candidates with AWS certification and 5+ years of experience."
• Provides the ability to compare multiple candidates side-by-side by simply asking, e.g., "Compare John and Priya for the frontend role."
• Made it easy to retrieve key hiring metrics like "How many candidates were rejected last month?" or "How many resumes were added this week?"
• Supports follow-up questions in natural flow, enabling deeper insights like "Why was this candidate rejected?" or "Do we have similar profiles?"
• Reduces dependency on technical teams for data access, empowering non-technical users to make faster, data-driven decisions.
• Integrates interview scheduling prompts into the chat, e.g., "Schedule an interview with John Doe."