Papa Owusu
A pen name, and a place to think out loud

Add value through tech.

I'm Papa Owusu. Solutions architect, application support analyst and data analyst by day. This is where I get to be curious in public: side projects, open source, and whatever tech I'm turning over this month. Lately that is mostly one question. How do we get more books and lessons into the languages people actually grew up speaking?

Follow @pkobts on X See what I'm working on
Illustrated portrait of Papa Owusu

The face is a drawing. The opinions are mine.

About

Papa Owusu is a pen name. The work under it is real.

My day job is solutions architecture, application support and data analysis. I like it. It is genuinely valuable work, solving real problems for real organisations. It is also proprietary and confidential, which is exactly as it should be, and it means there is very little of it I can talk about.

So this is the other half. Personal projects, open source contributions, and whatever technology has my attention this month, all completely decoupled from work. It keeps my skills current and gives me something I can actually share.

I am here to learn from people who are better at this than me, and to pass on anything from my own experience that saves someone else a bit of time.

Which is what the tagline is really about. Add value, in every direction. A post that saves you an hour. A reply that saves me a week. A conversation neither of us could have had on our own. Take anything here that is useful to you, and if you can improve it, even better. Tell me and I will learn something.

The pen name keeps the two lives apart so I can write freely. If you join the dots, no problem at all. Either way, my DMs on X are open.

pkobts
Papa Kwabena Owusu Behind The Scenes
Behind the scenes
Where the interesting, unfinished half of the work happens
Tagline
Add value through tech

Currently

Getting more books into low-resource languages

A "low-resource" language is one that computers have barely seen: little text online, few dictionaries, almost no training data. That includes languages spoken by tens of millions of people. The knock-on effect is quiet but heavy. Children learn to read fastest in the language they think in, and for a lot of them there is simply nothing printed in it. AI translation won't fix that alone, but for the first time it makes the job small enough for one person to start.

In progress

Translating with models, editing with people

Running public-domain and openly licensed texts through translation models, then working out where a human has to step in. Idiom, tone, names, anything culturally specific. The model gets you a draft, not a book.

I'm keeping notes on what breaks, because the failure cases are the useful part.

In progress

Publishing so it actually reaches someone

Translation is half the work. The other half is format, licence and distribution: getting a finished text into something a teacher can print, a parent can read on a cheap phone, and anyone can pass on legally.

Open licences by default. A book nobody can copy doesn't multiply.

Thinking about

Leaving the data behind

Every reviewed translation is a small pile of parallel text, exactly the raw material these languages are short of. Done carefully, and with the consent of the people whose language it is, the by-product may matter as much as the books.

Where I'm learning this from →
Not everything is that serious Word Remix is a word game that came out of a game night. It lives on the projects page with everything else I have built outside of work.
See all projects Play Word Remix

How I work

A few things I keep coming back to

Start smaller than feels serious

The projects I've finished all started as something slightly embarrassing: one script, one page, one chapter. The ambitious version was always available later.

Share the middle, not just the end

Polished write-ups teach less than "here's what I tried on Tuesday and why it didn't work." The middle is where someone else can join in.

Ask who it's actually for

If I can't name a real person who'd be glad the thing exists, it's a hobby. Which is fine, but worth being honest about before I spend three months on it.

Pointers

People doing this properly

I'm a newcomer to this field, not an authority in it. If the problem above interests you, these are better starting points than anything I've written. Mostly communities that have been at it for years.

Say hello

I'd rather this be a conversation

Tell me a language you'd like to see more books in, a tool that saved you an hour last week, or why you think I've got this wrong. All three are welcome.

@pkobts on X Day-to-day notes, half-formed ideas, and whatever I'm reading. Best place to reach me. github.com/pkobts Code and experiments, including whatever is behind this site.