Transcription to Job Requirement Matching using LLMs and NLP

Transcription to Job Requirement Matching using LLMs and NLP

The Challenge

An important part of the recruitment process is the interview. An inherently manual process, Tatras aimed to automate the reporting on interviews by analyzing candidate interview transcripts to improve consistency and accuracy in reporting of candidate fit. It allows for the identification of any inconsistencies or missing information that could impact the hiring decision, ultimately leading to a more thorough and reliable interview process. By addressing these gaps, organizations can enhance hiring decisions and a more robust recruitment process.

Hypothesis

  • The candidate interview call transcript is meticulously matched to the job description details.
  • The goal is to extract valuable insights from the conversation, including sentiment analysis, a concise summary, and identification of any gaps.

Execution

  • Candidate interview is provided as input with relevant job descriptions.
  • Interview transcript is shortened.
  • With different prompts tailored for different section and for many language models, this pipeline operates in stages
  • Saves results in MongoDB

Outcomes

  • Efficiency in summarising interviews and reducing the workload on interviewer to do it.
  • Takes 2-3 minutes to complete the overall process.
  • Can scale and fallback on models. Uses AWS step functions with lambda as span.

Project Highlights

manual and error prone process reduced
to 3 minute process with

95% Accuracy