The BLS projects roughly 129,200 software developer job openings per year through 2034, and almost every one of those roles will have a hiring manager clicking through at least one GitHub profile. Most of what they’ll find there won’t move the needle. Todo apps, weather dashboards pulled from a free API, e-commerce clones built along with a YouTube tutorial. They all blur together into the same gray heap. The question isn’t whether you should have a portfolio. It’s what goes in it.<\/p>\n
I’ve seen the same mistake play out repeatedly: a candidate spends three months churning out ten small projects, puts them all on their GitHub, and wonders why nobody responds. The issue isn’t quantity. It’s that nothing on the list answers the one question a hiring manager is actually asking: “Can this person build something that matters?”<\/p>\n
Hiring managers reviewing portfolios aren’t grading on effort. They’re pattern-matching. They’ve seen a thousand todo apps, and their mental model says “this person followed a tutorial” the moment one appears on screen. That’s not fair, but it’s real.<\/p>\n
The deeper problem is that tutorial projects don’t surface the things engineers actually get hired for: making trade-offs under uncertainty, reading existing messy code, shipping something to actual users who have opinions about it. A weather dashboard built in an afternoon can’t demonstrate any of that, even if the CSS looks great.<\/p>\n
The 2024 Stack Overflow Developer Survey<\/a> found that 66% of working developers hold a bachelor’s degree or higher, yet only 49% learned to code in a formal school setting. That gap tells you something: the credential matters less than most candidates assume. What fills the gap, for hiring managers evaluating candidates without traditional pedigrees, is evidence of real output. A GitHub profile full of clones doesn’t provide that evidence.<\/p>\n There are a few categories of projects that reliably separate portfolios that get callbacks from ones that don’t.<\/p>\n If something you built has 40 people using it regularly, you’ve had to think about error messages, edge cases, performance under real (if small) load, and what happens when someone does something you didn’t expect. That’s a fundamentally different experience from a project that only ever ran on your laptop.<\/p>\n Real users don’t have to mean a lot of users. A CLI tool you built that 30 engineers on Hacker News starred and actually use is more convincing than a full-stack SaaS app that’s never been deployed.<\/p>\n Building tools for other developers signals that you understand developer experience: you’ve thought about the person on the other side of the interface. A VS Code extension, a Homebrew package, a well-documented API wrapper: these are the kinds of things that engineers will actually download and try. And if they try it and find it useful, they remember the person who built it.<\/p>\n This category also tends to produce useful READMEs, because the audience demands them. A developer tool with poor documentation doesn’t get adopted. Forcing yourself to write clear instructions for other developers is one of the better writing exercises you can do before an interview.<\/p>\n The GitHub Octoverse 2024 report<\/a> counted 1.4 million developers making their first open source contribution that year. Getting a pull request merged into an established project (even a small bug fix or a documentation improvement) demonstrates something a personal project can’t: that you can read unfamiliar code, understand its conventions, and contribute without breaking anything. Those are core job skills.<\/p>\n If the PR is substantive (a real feature, a meaningful fix), it also gives you something specific to talk about in interviews: why you made certain decisions, what you had to learn, how the maintainer feedback changed your approach.<\/p>\n Scraping, cleaning, and visualizing data you found interesting is a strong signal for backend and full-stack roles. The key is that it should be original data and original questions. “I scraped 18 months of commit history from 50 repos I use and built a dashboard showing when maintainers are most active” is far more interesting than a project that just wraps an existing public API in a chart.<\/p>\n I’m going to be direct about something here: most candidates underweight the operational details and overweight the feature count. A project with CI\/CD through GitHub Actions, basic test coverage, and a clearly documented deployment process looks more production-ready than a feature-rich app with no tests and no way to verify it actually runs.<\/p>\n That last point is worth expanding. A README that says “why I made this” and “why I picked Postgres over SQLite for this use case” or “why I didn’t use Redux even though I could have” reveals a candidate who thinks about trade-offs. That’s a much stronger signal than a project with six features and no explanation.<\/p>\n A pattern we see in mock interviews<\/p>\n Candidates practicing on LastRoundAI often get tripped up when the interviewer asks “walk me through a hard decision you made in this project.” The candidates who struggle are almost always the ones whose projects were built from tutorials with no real trade-offs. The candidates who do well have a specific story: “I started with MongoDB and switched to Postgres halfway through because X.” You can’t fabricate that story if you didn’t actually live it.<\/p>\n<\/div>\n<\/div>\n The project itself is only half the problem. Presentation matters, and most portfolios fail here too.<\/p>\n A recruiter or hiring manager will spend 30 seconds on a project before deciding whether to dig in or move on. In those 30 seconds, they need to see what the project does, not what technologies it uses. “A tool that monitors my Postgres databases and sends me a Slack alert when query latency spikes above a threshold I set” is clearer than “a monitoring solution built with Node.js, Postgres, and the Slack API.”<\/p>\n A short screen recording (90 seconds, no voiceover required) demonstrating the app in a real scenario converts much better than a list of screenshots. If the project is a CLI tool, an animated GIF of a terminal session works well. The goal is to let a non-technical person understand what the thing does, and to let a technical person see immediately that it actually runs.<\/p>\n Two or three well-documented projects, each with a clear problem statement, a visible demo, and an honest explanation of what you’d do differently now, will outperform a portfolio of twenty clones. I’m fairly confident about that, though I’ll admit I don’t have data on how this advice changes for candidates applying specifically at very early-stage startups where breadth of exposure might matter more than depth.<\/p>\n Probably fewer than you think. For most mid-level roles, two strong projects are enough: one that shows you can build a backend service that handles real scenarios, and one that shows you can work on or contribute to something other people use. Senior roles are more interested in system design and decision-making than in whether you have seven projects or four.<\/p>\n The time you’d spend building a fifth forgettable CRUD app is almost certainly better spent going deeper on two good ones, or getting a PR merged into an open source project, or actually preparing for interviews through structured mock interview practice<\/a>.<\/p>\n Building good portfolio projects and being able to talk about them clearly are different skills. The second one is what most interviews are actually testing. If you’ve built something real but can’t explain your architecture decisions under pressure, you’ll lose to a candidate with a weaker portfolio who can.<\/p>\n If you’re preparing for coding interviews<\/a> or full-stack interviews<\/a>, practice articulating the project decisions you made. Not just what you built, but why, and what you’d change.<\/p>\nWhat actually stands out<\/h2>\n
Projects with real users, even a few dozen<\/h3>\n
Developer tools and CLI utilities<\/h3>\n
Open source contributions to existing codebases<\/h3>\n
Data projects with original analysis<\/h3>\n
Technical signals that matter more than you’d think<\/h2>\n
\n
How to present what you’ve built<\/h2>\n
How many projects do you actually need<\/h2>\n