AI agent for job hunting on hh.ru: vacancy monitoring and analysis in Telegram

Developer Ekaterina Kiyasheva spent several weeks actively searching for a job and found that the classic tools on hh.ru had stopped working. In response, she wrote an AI agent that takes monitoring into its own hands — and does it more effectively than the platform's standard newsletters.

What broke on hh.ru

The job market in 2025 is structured differently from how it was a few years ago. Opening a resume and waiting is a strategy of the past. Data from the platform itself shows: the peak of responses to a vacancy falls on the first day after publication. If a candidate applied a week later, the recruiter will no longer reach their application — the position is closed or attention has shifted to other candidates.

At the same time, the built-in search tools work unreliably:

• Auto-search resets some configured filters upon refresh
• Vacancies that clearly do not match the set parameters end up in the newsletter
• Relevant positions, on the contrary, often fall through the cracks
• Bugs cannot be reported: support works through a bot, with no direct connection to the developer
• Mass "open" applications yield no results — recruiters sense the absence of genuine interest

The result is an uncomfortable choice: either constantly check new listings manually, or send your resume to everyone. Both options either exhaust you or reduce the quality of your search.

How the agent works

Kiyasheva solved the problem in an engineering way: she wrote an AI assistant that automates routine monitoring. The agent runs on a schedule and goes through several sequential steps.

First, it scans hh.ru and collects vacancies published since the last check. It then applies filters: job title, tech stack, work format, region. At this stage, the majority of irrelevant positions are filtered out.

The remaining ones go through AI analysis: the agent reads the job description, studies the company page, and evaluates the requirements. The key step is matching the vacancy's requirements with the skills in the user's resume. The agent identifies what matches and what does not, and generates a fit score. This allows not just finding a suitable vacancy, but understanding how meaningful it is to spend time applying.

"I wrote an AI assistant that monitors the site on a schedule by

itself, selects only interesting vacancies, analyzes the company and the vacancy, matches the vacancy requirements with the skills in the resume, and delivers results to a personal chat."

The end of the chain is a Telegram notification with a curated selection of current vacancies and preliminary analytics for each. The user receives not just a link, but a brief summary: how well the position fits their experience, what the company is like, and what to pay attention to when applying.

How this differs from newsletters

Standard hh.ru notifications work on the principle of "keyword matched — add to list." The agent works differently: it evaluates a vacancy in the context of a specific person, not a faceless search query.

The user still decides for themselves whether to apply or not. The difference is that they do so consciously — with data in hand, not blindly. The tool is personalized: each person configures criteria for themselves and uploads their resume for matching.

What this means

Automating job searches is not a new idea, but previously it came down to mass applications. Kiyasheva's approach is fundamentally different: not "apply for you," but "help you choose consciously." As large platforms complicate their algorithms and lose transparency for job seekers, such personal filter agents become a practical tool — and more of them will keep appearing.