Staffing Search Tool - RAG/Function Calling

Staffing and recruiting companies are currently consigned to simple boolean searches on LinkedIn and going off of tags. Newfounding asked us to help solve this problem. My partner Zach Martin at Shaolin AI and I agreed and created this tool.

The goal

We wanted our users to be able to find a great match for a given search just through a natural language search. I need a programmer in austin texas should be enough to find you a programmer in that location. No messing with tags, geo filters, or anything else. You should type what you want and you get it.

Ingestion

We had a private dataset with docx and pdf resumes and some bios. We had to build out an ingestion pipeline for this to create vector embeddings for each member in their group. We opted out of chunking, because we wanted a vectorized search query to map 1 to 1 with a user. The results were pretty great without chunking so it never became a need. We did have an llm summarize the resume and bio before the embedding step to make sure we fit in the embedding context window. The whole ingestion pipeline looked like this.

Search

Once the data is ingested it allows for simple natural language searches like this.

Search Details

That powerful and simple user experience has a lengthy setup process to make it feel as good and be as simple and useful as it is. Here are the technical details.

End result

Our tool was able to find candidates that previously would have needed to be hand matched across many thousands of members. What was hours of work became seconds.

“Drew and Zach at Shaolin AI built the AI query system for our talent placement business. Unlike other platforms that were bloated and overly complex, their product was streamlined, performant and tailored to our needs. Highly recommend their work!”