Our Client:
A well-established recruitment agency that works with numerous companies across various industries
Situation:
The agency was facing a challenge in processing the vast number of CVs they received. Each client had different requirements for how they wanted the CVs formatted, and manually adjusting each CV was a time-consuming task. This approach is prone to human error and often taking up to 30 minutes per candidate. Additionally, the agency had to ensure that all CVs were anonymized and that pronouns were correctly adjusted to maintain consistency and neutrality, which ensures unbiased and fair treatment of candidates. This added another layer of complexity to the task. The agency needed a solution that could automate this process, ensuring each CV was correctly formatted and anonymized according to the specific client's requirements.
Task:
Our task was to develop a solution using Large Language Models (LLMs) to parse and format the CVs. The solution would take a CV uploaded by a candidate, process it using LLMs, and output a new document formatted according to the agency's client's requirements. We used LangChain for processing the CVs and Pinecone for storing the document embeddings, enhancing the efficiency and accuracy of information retrieval. We also provided an option to use either GPT-4 or GPT-3.5-turbo, allowing the agency to balance their needs for accuracy and cost-effectiveness.
Result:
The implementation of the LLM-based solution was a great success. The agency was able to process and format CVs quickly and accurately, taking only a few minutes per candidate to process with an additional few minutes to adjust minor details before sending them out to their clients. The client reported a noticeable increase in efficiency and a decrease in the time taken to process each CV. The dynamic parsing capability of LangChain, combined with the efficient retrieval of document embeddings stored in Pinecone, ensured that each CV was formatted according to the most recent and specific client requirements. The agency appreciated our innovative approach and tasked us with productionizing it with scalability and integration into their core system in mind.
