I built a custom LLM agent to handle my job search and it

A platform engineer at Zephyr built a custom LLM agent to automate his job search, using his public digital footprint to create a 'digital twin' that identified his niche and even recalled a March mention of an Anthropic role as his ideal position. The agent monitors job boards, filters for Spain/Remote roles, and suggests positions like Anthropic Technical Enablement Lead and DevRel at Databricks, while also assisting with interview prep and gap analysis.

I built a custom LLM agent to handle my job search and it Building a personalized LLM agent for recruitment The setup wasn't some complex coding project; I basically initialized a custom bot and gave it a clear persona. I didn't just say "find me a job." I gave it permission to scrape my public digital footprint to build a baseline profile. If you want to replicate this AI workflow, don't just feed it a static PDF resume. A resume is a curated lie; your public trail is your actual signal. Here is the prompt strategy I used to kick things off: Look at my online presence and find all the details you know about me. Use this to help me find my next role. Ask me any specific questions you need to refine the search. If you don't have a massive public footprint like I do, you should manually provide your LinkedIn URL or a detailed bio. The goal is to move away from generic keyword matching and toward semantic understanding of your career trajectory. The "Digital Twin" effect What blew me away was the model's ability to synthesize my history. It didn't just list my jobs; it identified a "through line." It saw my experience as a teacher, my work with the Playwright community at Microsoft, and my current role as a platform engineer on Zephyr's AI platform. It correctly identified my niche: teaching people how to build with MCP /en/tags/mcp/ and agents while actually shipping the workflows. Even more impressive was its ability to find "hidden" preferences. It dug through my old posts and found a specific mention from March where I identified an Anthropic role as my "ideal" position—with the caveat that I needed it to be remote due to family constraints. It connected that dot automatically. It wasn't just searching for "Platform Engineer"; it was searching for "the role I described as perfect three months ago." Real-world deployment and role discovery I've been running this as a continuous monitoring task. Instead of me checking job boards every morning, I just ask the agent for a status update. Search Parameters: I set it to prioritize Cursor /en/tags/cursor/ and Grok Bot, then expand to broader AI agent and education roles. Constraint Filtering: It automatically filters for "Spain/Remote" and flags US-based roles that might require an office exception. Aggregated Sourcing: It pulls from multiple different job boards and company career pages simultaneously, saving me hours of manual clicking. The results have been highly relevant. It suggested roles like Anthropic Technical Enablement Lead and DevRel positions at Databricks. It even provided a qualitative opinion: "If I were picking one exception to test, I would try Anthropic or Databricks." From discovery to interview prep The real power of this LLM agent comes when you move from "finding" to "digging." Once it identifies a lead, I don't just look at the title. I ask it to pull the full Job Description JD and analyze it. I can command the agent to: 1. Deconstruct the JD: Break down the specific technical requirements and soft skills. 2. Simulate Interviews: Use the JD and my profile to run mock interview sessions. 3. Gap Analysis: Identify where my current experience might fall short of the role's requirements so I can prepare specific talking points. This isn't just about automation; it's about having a high-level researcher working for you 24/7. If you're currently in the middle of a job hunt, stop treating LLMs like a search engine and start treating them like a specialized agent. Next Claude's 20-block cache lookback silently kills agent loops — → /en/threads/7182/ these real-world AI monetization case studies https://tanyan888.com/ , with plenty of directly applicable cases.