AI candidate sourcing pipeline with Cursor, Claude Code and Clay by Carolina FernandesAI candidate sourcing pipeline with Cursor, Claude Code and Clay by Carolina Fernandes

AI candidate sourcing pipeline with Cursor, Claude Code and Clay

Portfolio project
Carolina Fernandes
The problem A tech recruiting agency was finding AI engineers in the Netherlands mostly by hand, which meant going through conference speaker lists and looking people up one by one, and that takes a lot of time.
What I built
Python scripts, written with Cursor and Claude Code, collect speakers, talk titles and abstracts from public conference schedules, like the AGNTCon and MCPCon Europe 2026 one.
A second source finds people online through GitHub, Hugging Face, X, portfolios and public posts, and keeps them in their own Clay table with the same review steps.
An AI evidence search looks for proof of what each person has actually built.
A hiring rubric I wrote turns that into a verdict from strong to no, with a one-line summary and their most relevant project.
The review view puts the strongest candidates first, hides the nos and only shows the columns the person reviewing needs, and a Seen checkbox keeps track of who has been looked at.
How I designed it I approached it like a product, so I started by mapping how the sourcing worked by hand, found the slow steps and shaped the workflow around what the person reviewing actually needs to see. I also kept credit costs in check by filtering people early and only running the paid AI steps where they could change the verdict. The conference scripts only read the public schedule and never log in anywhere, and nothing goes to Clay until I run it on purpose and confirm.
The result Sourcing went from searching by hand to reviewing a shortlist with the strongest candidates at the top. Happy to walk you through how it works on a call.

0 likes · 2 views · Nov 17, 2026

I built an AI pipeline for a tech recruiting agency that finds AI engineers, checks what they’ve built and sorts them into a shortlist to review.

Tags