AI

A career knowledge base that builds tailored resumes where every bullet traces back to its source.

I was trying to solve a problem, as you do. My effort to make an LLM-powered job and resume management system, I realized that I needed to be able to pull apart a job description, figure out what it was really asking for, then match that somehow against a scattered career history to figure out what parts of my unique background were actually relevant to that specific job. I wanted to be able to tell, concretely, which piece or pieces of my history supported a particular requirement, and how. ...

The most dangerous thing that can happen with your resume when working with an LLM to optimize for job applications isn’t an outright lie. It’s a completely believable exaggeration. Led the decomposition of monolithic Java applications into modular Spring Boot services, defining service boundaries and communication patterns to support independent feature development. Sounds good, right? It does, except what really happened was more like Part of a six-person consulting team decomposing a legacy monolith for a large financial services client. I worked on service boundary decisions and some of the Spring Boot implementation. One of several engineers on the project. That bullet made it into exactly one application before I caught it. One. And I could have missed it even then. ...

The lying resumes had to stop. After enough customizations for enough job descriptions, my resume had started to look like a copy of a copy of a copy: roughly the right overall shape, but not quite recognizable as the same document. As I’ve been seeking a new job opportunity this time around, I’ve been trying to be selective about where I apply, but also to tailor the resume I submit to highlight how I fit the specifics of each job description. To avoid pulling my hair out, I have also started to rely on an LLM to help. The output typically meets my standards, but one major issue that I’ve started to notice is that my repeated modifications had started to introduce a lot of drift from ground truth and had lost the plot. For example, one position working with a large international bank customer had somewhere picked up the location of Toronto, Ontario. I’ve been to Toronto exactly once, and I never left the airport. ...