I'm working on a project where I need to analyze a large group of resumes stored in PDF format. The goal is to have Claude evaluate these files and provide summaries of the most suitable candidates, along with a scoring matrix based on specific metrics derived from the resumes. I've tried a few approaches so far:
1. First, I used a file management system to process around 30-50 resumes. Claude started reading only a few of them and created a report but didn't analyze all of the files, and after a while, it stopped responding.
2. Then, I attempted to upload the resumes directly to the project knowledge base and followed the same method. Unfortunately, I faced similar issues.
3. My last try was merging the PDFs into a single file and uploading that. This worked somewhat better, but I'm limited to processing only around 20-30 resumes at a time before encountering issues.
I've also tested other AI tools like Gemini and ChatGPT, but they all seem to struggle when dealing with more than 30-50 resumes. It makes me wonder if these tools aren't meant for this kind of task. I'm considering trying n8n or something similar. Does anyone have suggestions or methods that could work for analyzing multiple resumes effectively?
2 Answers
It sounds like you're trying to take on too much at once! Instead of feeding Claude all the resumes together, try processing them one at a time. You could score each resume individually and then pick out the top candidates for a more detailed review. This could help keep everything manageable and ensures you're getting accurate evaluations. Also, have you looked into using a script for headless Claude Code? That might help you tackle the larger batch without hitting limits.
Consider using a resume-specific parser API like Affinda or Sovren. These services can convert your PDFs into structured formats (like JSON) that include all the key details like education and skills. From there, you can use a tool like n8n to process each resume in bulk without running into the limits you're facing now. After parsing, you can then compute the metrics you need and get Claude to rank the top candidates based on these structured datasets instead!
That's a solid plan! Breaking it down into structured data first should make it way easier to manage and analyze all those resumes at once.
That makes sense! I feel like assigning scores to multiple resumes at once can really muddy the results too. Focusing on individual resumes and then narrowing it down sounds like the way to go.