pashwinkrish22/laya-email-resume-classifier

Laya (Fine-Tuned for Email Resume Classification)

This is Laya (convaiinnovations/laya) fine-tuned on a private dataset of job-application emails, to classify each email into one of four categories: freshers, experienced, referrals, others.

This model was trained on a private, anonymised dataset and is not evaluated against the public LocalLLaMA/typed-decisions benchmark.

Validation results

Accuracy: on a held-out validation split of the private dataset.

Installation & Quickstart

pip install laya

import laya

agent = laya.load("pashwinkrish22/laya-email-resume-classifier")

state = {"body": "...", "resume": "..."}
questions = {
    "category": {
        "type": "choice",
        "instructions": "Classify this job-application email.",
        "criteria": {
            "freshers": "recent graduate, 0-1 years experience, internships only",
            "experienced": "candidate with 1+ years of full-time work experience",
            "referrals": "candidate says they were referred by an employee",
            "others": "not a job application, or unclear",
        },
    }
}
result = agent.predict(state, questions)
print(result["answers"])

Data & privacy

Trained on internal job-application emails. Personal identifiers (names, emails, phone numbers) were scrubbed before training. This repo is private; do not make it public without review, since the base encoder can still memorise fragments of training text.

License

Apache 2.0 (base model: Convai Innovations' laya).