Sequence prep,
done in the kitchen.

We all have to sit and stare down sequences sometimes. A FASTA someone renamed three times. A FASTQ with suspicious quality tails. A handful of duplicate records that may or may not matter. A folder of isolates that need comparing before anyone can make a confident next move. None of it is hard. All of it is fiddly. And it was quietly eating my days.

I'm a phage biologist, so my real work is isolating phages, assembling genomes, and figuring out what's actually encoded in them. But every one of those days seemed to start with the same chores: dedup this, QC that, find the ORFs, compare these two, clean up a protein family before it goes into an alignment. The tools to do all that already exist and they're excellent — but they lived on my machine as a pile of one-off scripts and half-remembered command-line incantations, each with its own flags, its own output format, and its own opinion about what my input should look like.

Smart people should be working on hard problems that change the world — not the monotonous stuff that makes you want to pull your hair out.

That's the whole idea behind Genomics Kitchen. I wanted a place that quietly does the heavy lifting of prep so the scientist can spend their actual brain on the actual science. Open a page, drop in your files, get clean and reproducible results, close the tab. The five tools map exactly onto the chores I kept repeating: SeqSieve for dedup with full provenance, ReadLens for QC, SeqCompare for pulling apart core and accessory sets, ORFScout for honest six-frame ORF finding, and HMMForge to prep a protein family and hand me the exact MAFFT and HMMER commands to run next.

There was a quieter reason too. A lot of what I work with is unpublished, and uploading a sequence set to some random web tool just to get a quick quality plot is exactly the kind of thing you're not supposed to do — yet the convenient tools are so often the ones that phone home. So everything here is parsed in your browser and stays on your machine. Convenience without the exposure. That constraint isn't a limitation; it's the entire point.

I'll be honest about how it got built: I'm a scientist who's gotten good at building the tools I need with AI as a partner, not a career software engineer, and I'm not going to pretend otherwise. The ideas, the scientific framing, and the stubbornness about honest labels — exact matching called exact matching, ORF prediction never dressed up as annotation — those are mine. The implementation went a lot faster with help. Building it also forced me to understand the problem more deeply than using tools ever did: you can't hand-wave what "core" versus "accessory" means when the code actually has to produce the number.

Genomics Kitchen isn't finished, and it has gaps I know about and probably some I don't. But it runs, it's honest about what it does, and it's already made my own prep a lot less painful. For me, that's the right bar for whether a thing was worth building. If it saves you an afternoon of hair-pulling too, even better.