Skip the Complexity: How to Flatten Nested Data for End Users Fast

Perfect data modeling is a trap when someone is waiting on a working tool right now. This video is a small, concrete argument against letting that trap slow you down: a company dataset with 500,000 records gets flattened into roughly 50,000 usable rows in a matter of minutes, using a script generated by Claude, with the explicit decision to skip anything that complicates the schema without adding value for the immediate use case.

The problem with nested data

Company records in the source dataset include one-to-many relationships: multiple founders, multiple funders, multiple data points that don't fit neatly into a single row of a spreadsheet. The instinct for a careful data engineer is to model those relationships properly, with linked tables or nested structures that preserve every founder and every funding source. That instinct is correct for a mature product. It is the wrong priority when the actual goal is getting a searchable list into someone's hands today.

The deliberate shortcut

The fix here is intentionally blunt: if a company has seven founders, keep the first one and drop the rest. If it has three sources of funding, that level of detail isn't essential for someone trying to find a company to apply to. The flattening script, generated by asking Claude to build it, converts the nested JSON into a flat CSV built around the fields that actually matter for the immediate task: company name, industry, funding amount, city and state, phone, street address, year incorporated, filing date, and a primary contact.

Why this particular shortcut is defensible

The reasoning holds up because of who the data is for. This flattening work is part of the 80 Days to Stay initiative, which helps international students find visa sponsorship opportunities, and the resulting CSV lets someone search by state, filter by industry, sort by funding amount, and pull contact information immediately. A student trying to find companies to apply to in California or Massachusetts does not need to know a company had three funders; they need a name, a location, an industry, and a way to reach out. Running the script and checking the resulting industry stats shows the loss from flattening is minor, described as affecting less than 1% of the data.

Version 1.0, not the final version

The video is explicit that this is a first pass rather than a finished product. Planned next steps include adding annotations about companies' likelihood of hiring internationally, verifying career pages, detecting which applicant tracking systems companies use, and eventually building a crowdsourced verification system. None of that blocks release of the current CSV, which is described as ready to send to people who need it right now, in CSV, JSON, and Excel formats through the project's GitHub repository.

Key takeaways

• A 500,000-record nested JSON dataset was flattened to roughly 50,000 usable rows using a Claude-generated Python script.
• One-to-many relationships, like multiple founders or funders, were deliberately simplified rather than fully modeled, since they added complexity without immediate value.
• The resulting CSV focuses on fields a job seeker actually needs: company name, industry, funding, location, and contact info.
• This flattening effort supports the 80 Days to Stay initiative, which helps international students find visa sponsorship opportunities.
• The output is explicitly labeled version 1.0, with planned additions like hiring-likelihood scoring and career page verification to follow.

Try it yourself

If you're working on a data product for users with an urgent, immediate need, this video is a useful model: ship a simplified, honest version of the data now, and add sophistication once you've confirmed people are actually using it. The flattened dataset itself, in CSV, JSON, and Excel formats, is available through the linked GitHub repository for anyone supporting international students in their job search.