Developer Trained an AI Agent to Automatically Apply to Job Openings on HH.ru

A developer published a detailed breakdown on Habr of how they independently built an autonomous AI agent for job searching on HH.ru: the bot searches for vacancies on its own, filters out irrelevant offers using a local neural network, writes personalized cover letters for each position, and sends reports to Telegram — all without investing in paid APIs.

Why job search has turned into a war with robots

The author describes the IT labor market as merciless toward beginners: to get even one response, you need to respond to hundreds of vacancies, and most often you just receive automatic rejections from HR systems. On both sides of the process, algorithms now work: employers automatically filter resumes, and job seekers, to pass these filters, have to write a unique cover letter for each vacancy. This routine is exactly what the author decided to remove from themselves by entrusting it to their own agent, instead of manually rewriting the same text for hundreds of different job descriptions.

What the homemade AI agent can do

The agent's idea is simple: it works autonomously and completely handles the routine cycle of responding to vacancies while the user goes about their business.

• Independently searches for vacancies on HH.ru
• Filters out irrelevant and "trash" offers using a local neural network
• Generates personalized cover letters tailored to each vacancy
• Sends reports about found vacancies and responses to Telegram
• Works completely free, without paid APIs and extra configuration

Why a local neural network is needed

The author initially set a strict requirement: the script should be free and autonomous, without "song and dance" around paid APIs. Therefore, vacancy filtering — the most mass and routine part of the process — is performed by a local model that runs directly on the user's machine and requires neither subscription nor payment for tokens. This makes the agent completely independent from external services and their limits: no matter how many vacancies need to be processed in a day, the cost of running doesn't change and doesn't grow with search activity.

Cover letters, meanwhile, remain personalized for each specific vacancy rather than being mass-mailed with the same template — this, according to the author, should increase the chance of passing the automatic filters of HR systems, which easily recognize identical text sent to dozens of employers in a row. Telegram in this scheme serves as the only point of contact: the agent works in the background and sends reports about its work, not requiring the user to constantly monitor the process or manually run each stage.

How this agent differs from mass mailing of responses

The key difference in the approach is a rejection of simple spam responses. Many job seekers solve the vacancy quantity problem by simply mass-mailing the same resume and cover letter to everyone, but this approach fastest ends up in employers' automatic filters. The author took a different path: the agent first filters out irrelevant vacancies using a local neural network, and only then generates a letter oriented toward a specific position description. Due to this, the number of responses can be less than with mass mailing, but each one theoretically looks more meaningful from the perspective of a recruiter or HR bot reading it.

What this means

The story of a homemade AI agent for HH.ru is a characteristic example of how job seekers in an overheated IT labor market are taking automation tools into their own hands: since employers' algorithms are mass-filtering resumes, part of the candidates responds in kind and builds their own bots for responses and correspondence, not spending extra money or time on routines that used to be done manually.