arnav.raja(code, models, sound, colour, ideas)
Arnav
Raja
I write about the things I keep thinking about.
The shelf
Ten subjects, and the lines that run between them
Projects
Four branches of one domain, each split between what I make and what I take in
arnavraja.in / projects
-
lab.arnavraja.inIn progressBuilt
Large machine learning systems and full-scale experiments, with notes on what worked and what did not.
ReadPapers and books on how machines learn, with my notes on each.
-
tools.arnavraja.inIn progressBuilt
Ambitious, full-scale software I have built and made public.
UsedThe software I rely on every day, and what makes it good.
-
sound.arnavraja.inIn progressMade, less often
The occasional track, sketch or production note.
Heard, most of the timeListening logs, album notes and the records I keep going back to. This is most of what lives here.
-
colour.arnavraja.inIn progressMade, most of the time
Drawings, paintings and digital work. This is most of what lives here.
Seen, less oftenNow and then, notes on an artist, an exhibition or an image I keep returning to.
Also on this domain: blog.arnavraja.in for writing, and work.arnavraja.in for freelance work.
Writing
Ideas at three stages of ripeness
Writing on the blog is sorted by how far an idea has grown, not by date. Each piece carries a small copy of the shelf map, with the subjects it draws on lit up.
Seeds
Fragments, questions and things I noticed.
Sprouts
Notes that are growing: reading notes and ideas in progress.
Evergreens
Essays I have worked through and still stand behind.
Pieces move from one stage to the next as they grow. A seed can become an essay; most stay seeds.
Read the blog at blog.arnavraja.inWork
For people building things
Each job turns noise into signal. Something tangled comes in, and something clear goes out.
-
Software development
You bringAn idea, a process done by hand, or a tool that has outgrown its spreadsheet.
You getA website, web app or internal tool, built cleanly and handed over with documentation.
-
AI and machine learning
You bringA product, and a question about whether AI can make it better.
You getLanguage models, search or prediction added where they are worth using, and a straight answer where they are not.