The test
I took a real resume — a PES University CS grad, 7.1 CGPA, 2 internships, 4 personal projects, applying for ML engineer roles. I fed it to ChatGPT (GPT-4o) with the prompt "Analyze this resume and tell me how to improve it." Then I ran the same resume through ResumeCore.
Here's what each one said.
What ChatGPT said
ChatGPT scored the resume 82/100 and called it "a strong resume with good technical depth." It suggested:
- Add more quantification to bullet points ✓ (fair)
- Consider adding "orchestrated" and "spearheaded" to leadership descriptions (the resume had no leadership roles)
- "Your ML skills are strong — consider highlighting TensorFlow more prominently" (TensorFlow appeared in one bullet as "exposure to TensorFlow during coursework")
- "Overall this positions you well for ML engineer roles at top companies"
The problem: the resume had 3 months of internship experience and one ML project that used scikit-learn. It was not competitive for ML engineer roles at top companies. ChatGPT knew what the person wanted to hear, and it said that.
What ResumeCore said
ResumeCore scored the same resume 58/100. It said:
- ATS score: 71/100 — structure is clean but missing key ML keywords from target JDs
- Gap analysis: "Your ML experience is thin for the roles you're targeting. You have scikit-learn on one project. ML engineer roles at product companies typically expect TensorFlow/PyTorch in production contexts, not coursework exposure."
- Honest role fit: "You are competitive for junior backend SWE roles, junior data analyst roles, and ML roles at service companies. Not yet competitive for ML engineer at Flipkart/Swiggy/Zepto without additional project depth."
- Fabrication check: Flagged the "exposure to TensorFlow" — did not add it to skills or expand it
Why this matters
If you take ChatGPT's advice and put TensorFlow prominently in your skills, a technical interviewer will ask you about it in the first 5 minutes. When you can't answer, you've failed — not because you were unqualified, but because your resume overclaimed.
The honest 58/100 is more valuable than the flattering 82/100. It tells you exactly what to build before you apply again — one real TensorFlow project would push that score to 74 and make the ML role claim legitimate.
The broader point
AI tools that optimize for your positive reaction are not tools that help you get hired. They're tools that help you feel good about your resume for 10 minutes. The interview is a different story.
ResumeCore is built on the premise that honest feedback — even when it stings — is what actually moves your career forward. The 58/100 with a clear gap analysis is better than the 82/100 with a list of buzzwords to add.