Rolebot

Hiring has a strange bottleneck. The people best at judging candidates spend most of their time hunting for them, scrolling through profiles and sending messages that mostly go unanswered. Rolebot’s bet was that the searching could be automated, so a recruiter’s day could start with a list of good people already found, instead of an empty search box.

You tell Rolebot who you’re looking for. Each day it comes back with a curated list of candidates who fit, most of them people who weren’t actively job hunting. You pick the ones you like, Rolebot reaches out on your behalf and, when someone’s interested, the interview lands on your calendar. Teams could open new roles without hiring more recruiters, fill positions faster, and reach people they would never have found by hand.

I was one of the engineers on the team from 2023 to 2024, mostly on the backend and the infrastructure. That meant the services on AWS in Node.js and Python, the pipelines that sourced and ranked candidates overnight, the Elasticsearch layer that kept searching across millions of profiles fast, and the Laravel API the React app talked to. On the ops side I looked after deployments, environments and monitoring, and spent a lot of time making sure the nightly runs finished on time and didn’t cost more than they should. It was my first product where a machine learning model sat in the middle of the core workflow, and much of the job was making that reliable at scale.

Rolebot has since merged with Visage, another AI recruiting company, and the team, tools and platform now live under the Visage brand.