2 ms·

We just shipped portfolio-level screening in InterviewFlowAI

1 points by mukulmunjal 10mo ago

mukulmunjal 10mo ago
We noticed something odd while reviewing thousands of candidates: resumes were giving us an increasingly distorted picture of actual ability. People can list anything. Their real work is almost always sitting in a GitHub repo, personal site, Figma file, or some half-forgotten project link. So we built portfolio-level screening into InterviewFlowAI. When a candidate submits a portfolio URL, we analyze the actual project: – structure and depth of work – tech stack used (and whether it looks “real”) – clarity of writing and documentation – attention to detail – common red flags (broken links, empty repos, template sites, shallow descriptions) From that, the system generates: – a match score – strengths and gaps – a short summary the recruiter can skim in a few seconds Early tests show that portfolio signals correlate far more with real-world ability than anything in a resume, especially for engineers and designers. Curious if others here have seen the same shift: Are resumes becoming noise? Would love thoughts from people hiring at scale or candidates who prefer showcasing actual projects.
lionkor 10mo ago
So people who, for example, spend a lot of time maintaining a couple widely used legacy FOSS apps would lose to someone generating modern looking AI slop? I don't think you can win the screening process by automating it away. People are ultimately what drives every aspect of your org. Neglect the screening process by automating it away, and you'll end up with an org full of posers.
- mukulmunjal 10mo ago
  Totally fair concern, and I agree that "automate it away and forget about it" is a fast path to a team full of posers. Couple of clarifications on how we use this: • It is not “modern looking = good." A repo with years of steady commits on an "ugly” legacy app can score very high if there is depth, complexity, and real usage. • We explicitly look for things like long term maintenance, non trivial diffs, real world integrations, and documentation, not just shiny landing pages. • The output is a signal for a human, not a gate that replaces one. It helps a recruiter who sees 200 links decide which 10 are worth a closer look. In other words, it is meant to reduce noise, not replace judgment. If someone is quietly maintaining a critical FOSS project, they should stand out more, not less

1 points by mukulmunjal 10mo ago

We noticed something odd while reviewing thousands of candidates: resumes were giving us an increasingly distorted picture of actual ability. People can list anything. Their real work is almost always sitting in a GitHub repo, personal site, Figma file, or some half-forgotten project link. So we built portfolio-level screening into InterviewFlowAI. When a candidate submits a portfolio URL, we analyze the actual project: – structure and depth of work – tech stack used (and whether it looks “real”) – clarity of writing and documentation – attention to detail – common red flags (broken links, empty repos, template sites, shallow descriptions) From that, the system generates: – a match score – strengths and gaps – a short summary the recruiter can skim in a few seconds Early tests show that portfolio signals correlate far more with real-world ability than anything in a resume, especially for engineers and designers. Curious if others here have seen the same shift: Are resumes becoming noise? Would love thoughts from people hiring at scale or candidates who prefer showcasing actual projects.

lionkor 10mo ago

So people who, for example, spend a lot of time maintaining a couple widely used legacy FOSS apps would lose to someone generating modern looking AI slop? I don't think you can win the screening process by automating it away. People are ultimately what drives every aspect of your org. Neglect the screening process by automating it away, and you'll end up with an org full of posers.

mukulmunjal 10mo ago
Totally fair concern, and I agree that "automate it away and forget about it" is a fast path to a team full of posers. Couple of clarifications on how we use this: • It is not “modern looking = good." A repo with years of steady commits on an "ugly” legacy app can score very high if there is depth, complexity, and real usage. • We explicitly look for things like long term maintenance, non trivial diffs, real world integrations, and documentation, not just shiny landing pages. • The output is a signal for a human, not a gate that replaces one. It helps a recruiter who sees 200 links decide which 10 are worth a closer look. In other words, it is meant to reduce noise, not replace judgment. If someone is quietly maintaining a critical FOSS project, they should stand out more, not less