Automate job postings, track candidates, streamline interviews and offers, and review candidates efficiently.
job-eval.com
0Popularity
About job-eval.com
job-eval.com provides an explainable job fit scoring system using a 156M-parameter ModernBERT model. The tool evaluates how well a resume matches a job description by producing four separate sub-scores: requirement coverage, seniority alignment, domain fit, and ATS-style keyword overlap. Each score is accompanied by readable evidence paragraphs that show the model’s reasoning, allowing users to inspect and justify the results. The system can be used either through a live web interface or run locally on a user’s own hardware, including CPUs or GPUs. For technical users, the open-source code and model weights are available for batch processing entire job boards or applicant datasets. The model was benchmarked against a frontier LLM’s judgments on 100 resume–job pairs, achieving 0.83 quadratic-weighted kappa agreement and 99% of scores within one point of the frontier model’s ratings. The tool is designed for recruiters, hiring managers, job seekers, and engineers who need transparent, cost-effective resume screening without relying on external APIs or proprietary services.
Key features
- Four sub-scores: requirement coverage, seniority, domain alignment, keyword overlap
- Span-level attention converted into readable evidence paragraphs
- Local execution on CPU or GPU with no external API calls
- Batch processing of resumes and job descriptions via CSV input
- Open weights and MIT-licensed code available on GitHub and Hugging Face
- Web interface for single-resume evaluations with account-based history
- ModernBERT 156M-parameter model fine-tuned for resume-job fit scoring
- CSV output with scored results and explanation files for batch runs
Use cases
- Recruiters screening applicants with transparent, defensible fit scores
- Job seekers optimizing resumes by comparing against specific job descriptions
- Engineers automating resume screening for job boards or applicant tracking systems
Pros
- Open-source and MIT-licensed with no per-call costs
- Runs locally on CPU or GPU with no data leaving the user’s machine
- Provides four separate sub-scores plus explainable evidence paragraphs
- Benchmark-tested against a frontier LLM with high agreement scores
- Supports both web interface and self-hosted batch processing
Cons
- No cloud API or hosted service beyond the live evaluator
- Requires local hardware for batch processing of large datasets
- Limited to ModernBERT-based scoring without alternative models
Frequently asked questions about job-eval.com
What is job-eval.com and how does it work?
job-eval.com is an explainable job fit scoring system that uses a 156M-parameter ModernBERT model to evaluate how well a resume matches a job description. It produces four separate sub-scores—requirement coverage, seniority alignment, domain fit, and ATS-style keyword overlap—each accompanied by readable evidence paragraphs that show the model’s reasoning.
Who should use job-eval.com?
The tool is designed for recruiters, hiring managers, job seekers, and engineers who need transparent, cost-effective resume screening without relying on external APIs or proprietary services.
How can I use job-eval.com?
Users can either paste a resume and job description directly into the live web interface or run the open-source code and model locally on their own hardware, including CPUs or GPUs. The system offers two front doors: a web-based evaluator for non-technical users and a GitHub repository for technical users to automate batch processing.
Is job-eval.com free to use?
Yes, job-eval.com is free and open-source, with MIT-licensed code and open weights available on Hugging Face. There are no API keys, per-call costs, or lock-in requirements.
Can I run job-eval.com on my own hardware?
Yes, the tool is designed to run locally on a user’s own hardware, including CPUs or GPUs. The model is small enough to run on a laptop, and the open-source code allows for batch processing of entire job boards or applicant datasets without sending data externally.
How accurate is job-eval.com compared to frontier LLMs?
The model was benchmarked against a frontier LLM’s judgments on 100 resume–job pairs, achieving 0.83 quadratic-weighted kappa agreement and 99% of scores within one point of the frontier model’s ratings. It is designed to deliver frontier-quality results at a fraction of the cost and resource requirements.
