🚀 Alldone

Inspiration

Developers spend hours tailoring resumes for every job application by rewriting project descriptions, updating skills, and matching ATS keywords. We built Alldone to automate this process using GitHub as the source of truth.

What it does

Alldone analyzes a developer's GitHub profile and a job description to rank repositories, identify matched and missing skills, and generate an ATS-friendly LaTeX resume tailored to the target role.

How we built it

Built with Python and Streamlit, Alldone uses a hybrid pipeline combining deterministic extraction and LLM reasoning. It extracts technologies from GitHub metadata and dependency files, analyzes README files using Groq, Gemini, or Hugging Face models, ranks repositories by job relevance, and generates customized LaTeX resumes.

Challenges we ran into

• Reducing LLM hallucinations while keeping resume descriptions accurate.
• Handling GitHub API rate limits during repository analysis.
• Securely managing user-provided API keys in a BYOK architecture.

Accomplishments that we're proud of

• Built a hybrid deterministic + LLM repository analysis pipeline.
• Automated repository ranking and skill-gap detection.
• Generated ATS-ready LaTeX resumes directly from GitHub profiles.
• Supported multiple LLM providers with a unified workflow.

What we learned

We learned that combining deterministic extraction with LLM reasoning produces more reliable, transparent, and efficient AI systems than relying solely on LLMs.

What's next

• Perform deeper codebase analysis.
• Improve repository ranking with better semantic matching.
• Add more ATS-optimized resume templates.
• Deploy a scalable cloud-hosted version.

Built With

• agent
• api
• python
• request