Most government data portals are painful to use. They are slow, cluttered, and often require you to download a massive CSV file just to answer a simple question like, "What is the median salary for a Software Engineer in Austin?"
I built h1bsalarydata.org to solve exactly that friction. It is a lightweight, fast interface on top of the U.S. Department of Labor’s LCA data. No accounts required, no complex dashboards, just immediate answers about visa sponsor salaries and employer trends.
Here is the story behind why I built this and how it works under the hood.
The Problem with Raw Government Data
The underlying data source for this project is public domain: the Labor Condition Application (LCA) records filed by employers sponsoring H-1B workers. This data is incredibly useful for developers, hiring managers, and HR professionals trying to benchmark compensation or understand market rates.
However, the native interfaces provided by government agencies are often designed for bulk processing rather than quick human lookup. If you want to know the median wage for a "Data Scientist" in "Chicago" sponsored by "Company X," you typically have to:
- Download a CSV with hundreds of thousands of rows.
- Open it in Excel or Python.
- Filter by location, title, and employer.
- Calculate the median manually.
This workflow is robust but slow for quick checks. I wanted a tool that felt like a modern web app but served a strictly utilitarian purpose: answering salary questions in seconds.
How It Works: Simplicity Over Complexity
From an engineering perspective, this project is intentionally boring. That is a compliment. It relies on a straightforward architecture that prioritizes speed and reliability over flashy features.
The backend aggregates the raw LCA data and pre-calculates medians and counts for common combinations of job titles, cities, and employers. By pre-computing these aggregates, the site avoids heavy query loads during peak usage times. The frontend is a lightweight single-page application that fetches these pre-calculated stats and renders them instantly.
There are no complex caching layers or microservices here. The goal was to keep the time-to-first-byte low and the dependency tree minimal. When you search for a role, the site matches the input against the pre-indexed dataset and returns the median wage level, the number of applications, and the top sponsoring employers.
Why Median Matters More Than Average
One design choice I made was to focus heavily on medians rather than averages. In salary data, averages can be skewed heavily by outliers—such as a single executive earning $500k in a pool of junior developers. The median gives a much more honest representation of what a typical worker in that role actually earns.
The site breaks down salaries by Wage Levels (Level 1 through Level 4), which corresponds to the complexity of the job and the experience required. This allows users to see not just what people earn, but how compensation scales with seniority within specific companies and locations.
For example, if you are hiring a "Senior Backend Engineer" in Seattle, you can instantly see the median wage for Level 3 positions sponsored by major tech firms versus smaller agencies. This context is often missing from generic salary aggregators that blend all experience levels together.
Open Data, Closed Loop
The project is open-source and relies entirely on public data. There is no proprietary scraping or hidden data sources. Every number you see traces back to an actual LCA filing. This transparency is crucial for trust. When a candidate sees a salary range, they need to know it comes from real filings, not just self-reported surveys which can be biased toward higher earners.
By keeping the interface simple, I aimed to make the data accessible to everyone, not just data analysts. Whether you are negotiating an offer, planning a hiring budget, or researching market trends, the goal is to provide clarity with minimal friction.
What Would Help You Most?
I am constantly iterating on the filters and display options based on early feedback. Currently, the focus is on job titles and location. But I am curious about what additional dimensions would be most useful for your workflow.
Do you find it more helpful to filter by specific tech stacks (e.g., React vs. Vue) within the job titles, or is the broader occupational category sufficient for your needs?
Top comments (0)