FAQ

A lightweight, fast job board that aggregates software engineering internships and new grad positions from trusted sources across the web. No ads, no clutter—just fresh listings updated every 2 hours. See our Stats page for sources and details.

Every 2 hours.

Our job listings are sourced from popular GitHub repositories like 2027-SWE-College-Jobs and zshah101's automated internship list (MIT licensed), and directly from employers' public job boards (Greenhouse, Lever, Ashby, SmartRecruiters, Workable and Rippling), from which we keep software internships and new grad roles. We are looking to expand our sources to other community-curated lists that focus on SWE opportunities.

Once you sign in and add your resume, each listing can get a score from 1 to 100 for how well it fits you: 70+ is a strong fit (green), 40-69 a partial fit (amber), and under 40 a weak fit (grey). Click a score to see the breakdown behind it. Scores are private to your account; visitors who aren't signed in never see them.

The score is decomposed into five dimensions, each rated separately against your resume and preferences and then weighted: required skills 35%, experience level 20%, role similarity 20%, education 15%, domain 10%. Two checks act as gates: if the location, role type or timing clearly don't fit, or the posting lists a hard requirement you don't meet, the score is capped at 25. The ratings come from a calibrated decision model (TypeSafe's jev, via OpenRouter) that answers each question with probabilities over fixed, described levels. It is not an LLM chat and writes no text; the short reasons you see are generated from the numbers. Every job is rated twice, once with the job listed first and once with your profile first, and the two are averaged, so the order the model reads things in doesn't sway the result. When the two passes disagree, the breakdown says the confidence is low.

Every score is a paid call to the scoring model, and the site is free with no ads, so each person gets up to 60 new scores a day, and there is a site-wide daily spending cap. When you hit either, the score shows limit; it resets at 00:00 UTC. Scores you already have stay, and re-opening them is free.

We fetch each posting's full description from the employer's site. When that fails (the page is behind a login, blocks automated requests, or was taken down), the job is scored from its title, company, location and type alone. Those scores are shown muted with a ~ in front: treat them as a rough estimate.

A score is tied to the exact inputs it was computed from: your resume, your preferences, the job's description and the model version. When any of them changes (you upload a new resume, edit your preferences, or we fetch a description for a job that was scored on its title), the old score is out of date and the job is scored again. Adjustments to the weights can also move scores slightly.

We keep only the text of your resume (the uploaded file is read in memory and never saved), with your preferences, on your account. To compute a score, that text and your preferences are sent with the job's description to OpenRouter, which passes them to TypeSafe's scoring model; they are used only for that. We never sell them or share them with employers. You can replace your resume or delete your account, resume and scores at any time from your account page. See the Privacy Policy.

Each listing gets an automated estimate of whether the employer will sponsor an H-1B visa for the role, made by a model from the posting text (or, when we couldn't fetch the description, from the title and company): yes means the posting says it sponsors, no means it rules sponsorship out or requires citizenship, a green card or a clearance, and likely / unlikely are guesses from the employer, industry and role when the posting says nothing. Hover a label to see whether it was stated in the posting or inferred. It is an automated estimate from the posting text, not legal advice: it can be wrong, so confirm with the employer before you apply.

It is an automated estimate from the same kind of decision model, made from the posting text plus whether the employer appears in public USCIS H-1B employer data. A label is stated when the posting itself says something about sponsorship or work authorization (solid outline), and inferred when it says nothing and the label is a guess from the employer, industry and role (dashed outline). Stated labels are much more reliable than inferred ones, and labels for jobs without a fetched description are the roughest. It is not legal advice: always confirm with the employer.

We use Google Analytics to understand user behavior. We believe this is one of the least intrusive ways to keep track of our site's performance. For more info, see our Privacy Policy.