You’ve probably noticed it by now. You’re deep into an ML or robotics project, and you just need to quickly inspect a dataset. But your data is a mess of HDF5, Parquet, and Zarr files. You love the myHDF5 viewer, but what about the rest? You end up juggling three different web apps, each with its own quirks, just to glance at your own data. Or worse, you’re uploading sensitive datasets to some random cloud viewer just to see what’s inside.
The biggest threat to your data isn’t a breach; it’s the friction of trying to look at it.
We’ve been conditioned to believe that working with complex data formats requires complex, fragmented workflows. You have a tool for this, a site for that. But this fragmentation isn’t just annoying—it’s actively destroying your focus. Every time you switch tabs to inspect a different file format, you’re paying a cognitive tax. You lose your train of thought, you lose your momentum, and you lose the very insight you were chasing.
An ML/Robotics researcher recently got fed up with this exact problem. They loved myHDF5, but found no good equivalent for Parquet and Zarr. Switching between different sites was killing their productivity. So, they built ViewKit.
It’s an interactive dataset viewer that runs right in your browser. No uploads. No cloud syncing. No sending your proprietary robotics data to a third-party server just to render a quick visualization. It handles HDF5, Parquet, and Zarr files locally, seamlessly.
Every time you switch tools to inspect a different file format, you’re paying a cognitive tax that drains your focus.
Here is the twist: we thought the barrier to a unified data workflow was technical difficulty. We thought building a local, universal viewer was too hard. It turns out, the real barrier was just the mental overhead of accepting the status quo. We accepted that data inspection had to be a fragmented, privacy-compromising chore.
ViewKit proves that a simple, no-upload browser-based solution can eliminate that friction entirely. It bridges the gap between privacy and convenience—a gap that no existing tool satisfactorily crossed until now.
Privacy shouldn’t require a sacrifice of convenience.
If you work with these files daily, you know the anxiety of uploading sensitive data. You know the frustration of context-switching. ViewKit doesn’t just save you time; it restores your focus to where it belongs: on the actual analysis. Stop juggling viewers. Stop uploading your data. Just look at it.
FAQ
Q: A browser-based viewer can't handle large datasets as well as a native app, right?
A: For quick, interactive inspection—which is what you do 90% of the time—running locally in the browser eliminates the upload bottleneck and gets out of your way.
Q: So I can just open my files without an internet connection?
A: Yes. Since it runs locally and requires no uploads, your data never leaves your machine. You get the convenience of a web UI with the security of a desktop app.
Q: Is this really necessary if I can just write a Python script to inspect files?
A: You could, but writing a script every time you want to glance at a dataset is exactly the kind of friction that kills your momentum. Sometimes you just need to look at the data, not code a viewer for it.