Role overview
Software engineering is one of the highest-stakes hires a small company makes — the cost of a bad fit is months of slowed velocity. It is also the role where resumes are most misleading. Confidence, certifications, and prestigious employers correlate weakly with shipping output.
Adjust this scorecard hard for the level you are hiring (Junior / Mid / Senior / Staff). The signals shift dramatically across them.
Job criteria
- Stack overlap with your codebase (be specific about what cannot be wrong)
- Years of relevant experience — but read it as shipping years, not employed years
- Evidence of end-to-end ownership (designed, built, shipped, owned in production)
- Code samples — public repo, technical blog post, in-depth project description
- Comfort communicating in writing (PRs, docs, design docs)
- For senior+: track record of unblocking other engineers, design-doc authorship, mentorship
What to screen for
- Depth over breadth. A resume listing 30 technologies usually means surface knowledge of all of them.
- Ownership signal. "Built and shipped X to N users" is a different signal from "contributed to X."
- Public artifacts. GitHub, blog, conference talks, OSS contributions. Their absence is not disqualifying; their presence is signal.
- Shipping years vs employed years. A 5-year resume with two real shipped projects is different from one with twelve.
- Writing quality. PRs and design docs are most of senior engineering. The resume itself is the smallest writing sample you will get.
Red flags
- "Senior" title from companies where they were <2 years in
- Endless certification lists with no shipped projects
- Tech stack drift every job — no apparent specialization arc
- Buzzword soup ("microservices, Kubernetes, AI/ML, blockchain") without any of it tied to a specific outcome
- Resume claims that contradict the public GitHub
- Generic, AI-flavored cover letter that does not name your stack or product
Resume keywords
For a typical product engineering role: TypeScript, React, Node, Python, Go, Rust, PostgreSQL, Kafka, Redis, AWS, GCP, Kubernetes, Docker, CI/CD, production, shipped, architected, migrated, designed, code review, mentored, on-call
Replace this list with your actual stack. Asking the AI to score against a list that does not match your codebase is noise.
Interview questions
- Walk me through a system you designed and built end-to-end. Why those choices?
- Tell me about the worst production bug you have caused or debugged. What did you learn?
- What is something in your stack that you used to think was good engineering and now think is wrong?
- How would you decide between and for <decision relevant to your codebase>?
- Describe a piece of code you wrote that you are still proud of, and a piece you would now rewrite.
- (Senior+) Tell me about a design doc that changed direction after review. What was the original plan, and what shipped?