Writing

How I joined a classmate's open-source project and spent a few days closing the feature gap with the original Windows app.

What my team and I learned trying to decide where Ghana's bed nets should go when the data is thin, skewed, and missing in the places that matter most.

Building an n-gram language model for Ewe from scratch, testing when n-grams beat neural models, and adapting small LLMs to health and agriculture.

Over five days, four of us built and shipped an offline-first app for registering farmers in rural Ghana.

Over four weeks and about $300 in rented GPU time, I ran ten rounds of experiments on Orb's local question-answering pipeline. The biggest gain came from a reranker bug, not a strategy.

My heart aches

I wish that day didn’t happen

They say everything happens for a reason,

I am here dreaming about my good life with you.