Right here’s a pattern of the JSON the mannequin returns
{
“match_score”: 70,
“abstract”: “Neda’s resume demonstrates related expertise in buyer relationship administration, stakeholder engagement, and information evaluation…”,
“improvement_tips”: [
“Quantify achievements in the Real Estate Agent role…”,
“Tailor the resume summary to better address high-volume accounts…”,
“Emphasize experience with CRM tools like Salesforce or HubSpot…”
]
}
I stored the output structured so it may simply be utilized in an app or built-in into an internet platform.
Apart from the pure language output, I additionally used Gemini’s embedding mannequin to measure how shut the resume and job publish are — mathematically — utilizing cosine similarity. This provides a second perspective to validate how effectively the 2 align.
This type of software might be tremendous useful for:
- Job seekers who need suggestions earlier than they apply
- Profession coaches serving to purchasers optimize resumes
- HR groups screening candidates extra effectively
This was my first time working hands-on with Google’s Gemini SDK, and I realized rather a lot. Getting the structured output proper with immediate engineering was simple, however embedding assist took a bit extra trial and error. Particularly on Kaggle, I had to determine methods to correctly set up dependencies and name the precise API strategies for textual content embedding.
I additionally realized methods to assume like a consumer — not only a coder. This venture wasn’t nearly AI; it was about creating one thing useful, quick, and sensible.
👉 Click here to view the full project on Kaggle
I beloved engaged on this venture as a result of it felt actual — one thing that would assist individuals in a significant approach. There’s much more I’d love so as to add, like saving outcomes, batch evaluations, and even pairing this with a resume generator. However for now, it’s a strong basis I’m pleased with.
