我打造 HirePilot：一款本機運行的開源 AI 求職助手

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

What if preparing for a job interview didn't mean juggling five different tools?

Meet HirePilot AI 🚀 — an open-source, local-first AI career copilot that helps candidates go from a job description to an interview preparation plan.

Here's the workflow:

📄 Upload your resume + paste the job description. HirePilot analyzes your existing experience, identifies relevant skills, and highlights gaps against the role.

💬 Answer a few questions. When important details are missing, it asks contextual follow-up questions to improve your resume without inventing experience or achievements.

✍️ Generate an ATS-friendly resume. Get a role-tailored, editable resume with relevant keywords, plus PDF and DOCX exports.

🔎 Research the company. Enter a company name to discover relevant company information and publicly reported interview questions, with source links.

🧠 Prepare for the interview. Get role-specific technical and behavioral questions, suggested answers, and a personalized list of topics to revise.

✅ Track your readiness. Work through a prioritized preparation checklist and watch your progress move from 0% to 100%.

The idea is simple: one job description → one tailored resume → a complete interview preparation roadmap.

Who would pay for it?

Students, fresh graduates, job seekers, and professionals applying to multiple roles could use HirePilot to reduce repetitive preparation work.

A potential business model is a free local version with premium resume templates and additional career workflows. The goal is to make useful career preparation accessible without requiring a paid LLM subscription.

Code

The code is open source, and you're welcome to explore, fork, and build on it!

What's under the hood?

• 🧠 Gemma 3 4B + Ollama — local inference for resume analysis, tailoring, and interview preparation.
• ⚛️ React 19 + TypeScript + Vite — the frontend.
• ⚡ FastAPI + Python — the backend.
• 🔍 DuckDuckGo Search + optional Tavily — company research and sourced interview insights.
• 📋 Local progress tracking — persistent application sessions, resume edits, and preparation checklists.

The project is released under the MIT License.

How I Built It

The central idea behind HirePilot is that your career data shouldn't need to leave your computer just to help you prepare for a job.

I built the application around Google's Gemma 3 4B model, running locally through Ollama. Instead of relying on a proprietary cloud LLM API, HirePilot performs its core AI tasks using local inference.

The stack:

• AI: Gemma 3 4B through Ollama.
• Frontend: React 19, TypeScript, Vite, Tailwind CSS, and Lucide React.
• Backend: Python, FastAPI, Pydantic, and HTTPX.
• Resume processing: pypdf and python-docx.
• Company research: ddgs for DuckDuckGo search, with optional Tavily integration.
• Persistence: Browser localStorage for application sessions, edits, and preparation progress.

I separated the workflow into focused stages: resume and JD analysis, contextual clarification, resume generation, company research, interview preparation, and checklist generation.

The resume matching score is calculated from matching skills, requirement gaps, and keyword coverage. It is an estimate of alignment with the JD, not a guarantee of passing an ATS.

The application also distinguishes candidate-confirmed information from missing skills, so the AI doesn't need to fabricate experience to make a resume appear stronger.

Why Does Open Innovation Matter?

For HirePilot, open innovation is what makes a privacy-focused, affordable career tool practical.

🔒 Privacy by design

Resumes contain personal information and employment history. Running inference locally means candidates can use AI without sending their resume to a commercial LLM provider.

Company research is handled separately. Search queries contain relevant company and role information, not the candidate's resume or personal contact details.

💸 No mandatory per-request LLM fees

Once Ollama and the model are installed, the core AI workflow can run on the user's hardware without paying a commercial LLM API for each request.

There is still a hardware requirement, and web research needs an internet connection.

🔄 Freedom to experiment

Ollama makes it straightforward to experiment with different locally installed models. Developers can adjust prompts, generation settings, and model choices without tying the application to one proprietary API.

🌱 Accessible career preparation

A paid AI subscription shouldn't be a prerequisite for building a better resume or preparing for an interview. Local inference makes the core workflow more accessible while giving users greater control over their data.

That's the open-source advantage I wanted to explore: not just using an open model, but building a product around local inference, privacy, and user control.

Prize Categories

• Best Use of Gemma — HirePilot uses Gemma 3 4B as its core local inference model for resume analysis, tailoring, and interview preparation.