How an AI Job-Application Agent Cut Job Search Time by 90%

A developer published a detailed breakdown on Habr of building a fully autonomous AI agent that automates job hunting on HH.ru, Russia's largest job board. The project — built entirely with free, local tools — finds relevant vacancies, filters out noise with a local LLM, writes personalized cover letters, and reports back over Telegram, all without a single paid API call.

What Happened: An AI Agent That Applies for Jobs on Its Own

The author, frustrated by paid auto-apply services charging 3,000-5,000 rubles a month, built a free alternative: an asyncio-based Python bot using Playwright to bypass HH.ru's bot protection (Cloudflare, dynamic JS) by behaving like a real logged-in user. A local Llama 3 model running through Ollama filters out irrelevant postings — jobs for managers, HR, or unrelated roles are rejected automatically — and writes a unique, tailored cover letter for every match. SQLite tracks which vacancies were already processed to avoid duplicate applications, and Aiogram sends live status updates to Telegram.

Why It Matters: Leveling the Playing Field for Job Seekers

The core value isn't just automation — it's access. Junior developers and career switchers routinely lose to keyword-matching HR bots before a human ever reads their resume. An agent that mass-produces genuinely tailored applications removes that first-round filter disadvantage, and because the entire stack is free and open, there's no subscription fee gatekeeping the tool.

Impact on the Job Market

In competitive markets like IT, where a single posting can draw hundreds of applicants, tools like this shift the bottleneck from 'who applies fastest' to 'who has the best-matched profile' — a more meritocratic filter, assuming cover letters stay genuinely personalized rather than spammy.

How to Apply This Today

The architecture — a headless browser for scraping, a local LLM for filtering and generation, and a lightweight local database for state — generalizes well beyond job hunting: lead generation, market research, or partnership outreach can all follow the same four-module pattern (config, bot, AI analyzer, browser client).