Datasetpublic

Builder-Neekhil/career-agent-dataset-v1

Career Agent Fine-Tuning Dataset A curated multi-turn conversational dataset for training a personal career agent. Sources Source Examples Purpose cnamuangtoun/resume-job-description-fit ~4.6K Job fit assessment (JSON output) opensporks/resumes ~7.2K Resume review, interview prep, career paths sandeeppanem/resume-json-extraction-5k ~4.9K Structured resume parsing Synthetic coaching 600 Salary, pivot, networking, gaps, promotion Total: ~12K+… See the full description on the dataset page: https://huggingface.co/datasets/Builder-Neekhil/career-agent-dataset-v1.

sourceHugging Faceupdated 5mo agoView on Hugging Face

1likes38downloads

Dataset Card

Career Agent Fine-Tuning Dataset

A curated multi-turn conversational dataset for training a personal career agent.

Sources

Source	Examples	Purpose
cnamuangtoun/resume-job-description-fit	~4.6K	Job fit assessment (JSON output)
opensporks/resumes	~7.2K	Resume review, interview prep, career paths
sandeeppanem/resume-json-extraction-5k	~4.9K	Structured resume parsing
Synthetic coaching	600	Salary, pivot, networking, gaps, promotion

Total: ~12K+ multi-turn conversations (system → user → assistant)

Format

json

{
  "messages": [
    {"role": "system", "content": "You are a seasoned career advising expert..."},
    {"role": "user", "content": "Please review my resume..."},
    {"role": "assistant", "content": "## Resume Review Feedback..."}
  ]
}

Compatible with TRL SFTTrainer with assistant_only_loss=True.

Build it locally

bash

pip install datasets
python build_dataset.py

This will download source datasets, transform them into conversations, and push Builder-Neekhil/career-agent-dataset-v1 to the Hub.

Use it directly after building

python

from datasets import load_dataset
ds = load_dataset("Builder-Neekhil/career-agent-dataset-v1", split="train")