GoodJob: a BERT-based job post classifier
2 min readupdated
The GoodJob model (110M parameters, built on bert-base-uncased) is designed to analyze posts from Twitter/X or LinkedIn. It features two classification heads: the first determines the post type (ranging from specialized AI/ML job listings to non-work posts), and the second evaluates author behavior (a personal account versus a job board channel). This tool is useful for building data collection pipelines, filtering out noise, and finding genuine job offers.
Key features
- Two classification heads for simultaneous analysis of post type and account type.
- Precise categorization into ML Job, Job, Other, and Not Job.
- Behavioral account analysis to identify sources of mass job postings.
- Integration with the transformers library for quick code execution.
What you can do
- Set up automated collection and filtering of job openings for digests or aggregators.
- Clean social media feeds of informational noise before analyzing the job market.
- Automate the search for relevant orders and projects for developers.
from transformers import AutoModelForSequenceClassification
model = AutoModelForSequenceClassification.from_pretrained(
"SpiderHomie/goodjob",
trust_remote_code=True,
)
# Анализ одного поста
model.predict("We're hiring an ML engineer, DM me")
Example
from transformers import AutoModelForSequenceClassification
model = AutoModelForSequenceClassification.from_pretrained(
"SpiderHomie/goodjob",
trust_remote_code=True,
)
# Анализ одного поста
model.predict("We're hiring an ML engineer, DM me")
Read more
DetailsBenchmarks (test split of 483 posts): type head accuracy is 0.872, macro F1 is 0.854; macro F1 for the account head (board head) is 0.736.
Hardware requirements: standard configuration for the BERT family of models (around 110M parameters), with device="auto" support.
Limitations: maximum text length (max_length) is 256 tokens; the model was trained on a specific dataset from Twitter/X and LinkedIn and is not intended for hiring decisions.










Comments