I build scalable systems and AI applications. Currently exploring what aerial footage can teach a computer.
I work across the full stack of modern software, from the database up to the model serving layer, with a particular focus on systems that need to be both intelligent and reliable.
Designing and building secured distributed systems that survive contact with real users, real load, and real-world networks.
Production ML pipelines, computer vision, and LLM applications that ship, not stay in notebooks.
Long-form essays and tutorials that explain hard ideas in plain language. Currently writing Pixels from Above.
Capturing aerial footage across Ghana and turning it into datasets for computer vision tools.
A featured project and a few others in flight. I prefer to show working software over slide decks.
A browser-based computer vision platform for aerial footage. Detect, track and segment objects from drone video, all running locally in the browser via Pyodide. Built for analysts, surveyors, and curious pilots who’d rather not upload their footage to the cloud.
A flight-log + media organizer for drone pilots. Auto-tags footage with GPS, weather, and detected subjects.
A tiny Python parser for DJI flight-log files. Turns CSVs into pandas DataFrames with one call.
Building aerial computer vision tools, one article at a time.
Deploying large language models on mobile and IoT devices requires careful quantization and pruning. Here is a breakdown of techniques to reduce latency without sacrificing accuracy.
As generative models become ubiquitous, we must address the bias inherent in training data. A look into the societal impact of AI.
Not a list of everything I’ve ever touched; only what I’d willingly use again on a Monday morning.
I’m open to interesting work: contracts, collaborations, or just a coffee about computer vision. I usually reply within a day.