Generating Synthetic Job Ads through LLMs

This repo contains code for JobGen. If you like this paper, please cite us:

• The JobGen code
• It's required inputs in data/
• Notebook for metrics & experiments, in experiments/
• Results of previous Jobset, in results/

I used a vLLM server (https://docs.vllm.ai/en/latest/getting_started/quickstart.html) as inference engine. My recommendation is to install vLLM and run the server with this command python -m vllm.entrypoints.openai.api_server --model casperhansen/llama-3-70b-instruct-awq --quantization awq --gpu-memory-utilization 0.95 --enable-prefix-caching --enforce-eager for maximum performance. Then, use the following command to install dependencies:

pip install -r requirements.txt

At this point, we used to launch two concurrenct worker in order to fully utilize a continuos batching

python main.py 1
python main.py 2

In main.py set use_online_oja = False if you want to provide your own custom seed (real-life vacancies) otherwise the system will scrape it from the internet Also set up where do you want to store the generated OJAs. We used an AWS S3 bucket so you need to provide your access key and bucket name