JobPilot AI

Inspiration

Looking for a job has become a frustrating, time-consuming full-time job in itself. Candidates spend dozens of hours a week scrolling through fragmented job boards, manually tailoring their CVs for Applicant Tracking Systems (ATS), and filling out repetitive forms. We wanted to build a solution where the user does absolutely nothing but talk—leaving the heavy lifting to a truly autonomous agent.

What it does

JobPilot AI is an end-to-end autonomous job-hunting agent. The user simply logs in and has a natural conversation (via voice or text) with the AI about their experience, skills, and ideal role. From that single interaction, the agent takes full control: it scans the web for matching openings, reverse-engineers job descriptions, automatically optimizes the user's resume for each specific role, and executes the application process autonomously.

How we built it

The core of the system relies entirely on the Gemini ecosystem, utilizing its advanced reasoning and long-context capabilities across three main layers:

• User Onboarding (Gemini Pro): Processes the initial conversational interview along with the user's raw CV, portfolios, and history. Gemini’s massive context window allows it to deeply understand the user's professional background without losing detail.
• Smart Matchmaking (Gemini Flash): Connected to job market APIs and web scrapers, Gemini reads job descriptions like a human recruiter, evaluating cultural fit and technical alignment far beyond basic keyword matching.
• Autonomous Action (Function Calling & Tool Use): Gemini drives the automation. It dynamically generates custom, ATS-friendly CV variations and tailored cover letters for each job. Using Function Calling, it interacts with application APIs and web workflows to submit applications on the user's behalf.

Challenges we ran into

Building a reliable autonomous agent requires strict guardrails. Ensuring the AI accurately maps complex career histories to dynamic job descriptions without "hallucinating" qualifications was a major hurdle. We solved this by using structured JSON outputs and rigorous validation prompts within the Gemini API to guarantee alignment.

Accomplishments that we're proud of

We designed a seamless pipeline where a multi-step, complex real-world workflow (searching, filtering, rewriting, and applying) is condensed into a single, effortless user action: just talking to an AI.

What we learned

We learned how to maximize Gemini's Function Calling efficiency to orchestrate complex external browser workflows. We also discovered that leveraging Gemini's long context window to analyze a user's entire professional history yields far more accurate ATS-optimized resumes than traditional, template-based approaches.

What's next for JobPilot AI

We plan to expand the agent's capabilities to include automated follow-ups (sending thank-you emails to recruiters via Gmail integration) and an AI Interview Simulator that trains the user specifically for the jobs the agent has successfully secured for them.

Built With

• autonomous-agents
• function-calling
• gemini-api
• google-gemini
• javascript
• node.js
• python