My Experience at the Ghana Data Science Summit - IndabaX 2026: Data First

Last week I was in Ho in Ghana’s Volta Region for the Ghana Data Science Summit (IndabaX 2026). I was a speaker at the first IndabaX in 2019, when I had just started out in Machine Learning, so looking back, that was more bravery than I gave myself credit for. Anyway, it was nice to be back home. This was my first major conference since returning to Ghana last year.

I was initially going to try putting together a workshop, but I couldn’t assmeble the avengers in time, so I had to participate in other ways… i.e. reviewing workshops and joining the event’s hackathon.

So What’s New?

The faces! So many new people! And a lot of them are doing really cool stuff. I can’t possibly cover everything, but these are the experiences that stood out to me based on my work and interest these days, and my experiences during the conference. Also I missed the first day of the conference since I spent most of the dya on the road, so, at best, my experience covers just two-thirds of the timeline. It will have to do.

Everybody Loves Data

Everybody knows (or should know) that the best value you get from conferences is in the conversations in the hallways… that’s how actual connections are built. Also, there are no slides or GitHub links to be shared for review later, just contacts. So shoutout to all the cool people I met. If you’d like to access the tutorial material, they can all be found here.

These interactionsa are how I stumbled across a group of people discussing R. I hate R 🫩. However, it was a data science conference, not an R conference, so we could still talk around data. The theme for this year was, after all, “Data First: Unlocking Ethical & Inclusive AI in Ghana”. Imagine my surprise when I heard the wildest story of how one member of this discussion, a retired professor, had found Ghanaian (or in this case, Gold Coast) population datasets going back to the 1800’s somewhere in Austria, and had photocopies sitting at home beacause he couldn’t reliably digitize them. Well, keep your eye no this space. If you hear that I’ve released such a dataset, you’ll know how it all came about.

The other dataset conversation that really interested me came through the Open Data Bank project (shoutout to Afua Anaglate and team!). This has to do with NLP for low-resource languages (whicih should be better-renamed “low-digital-resource” languages or something like that). This interested me because of prior context at IPA, and issues that exist around data extraction from local languages during surveys. I had also spent the week before discussing issues with African-accented English with a friend, and working on LingoShadow which showed me the gotchas that pop-up with non-native English. So here’s the TL;DR, you can contribute your language data on Mozilla CommonVoice, and here’s a Google Colab Notebook where you can see ASR fine-tuning in action. None of this is new, but let’s consider this extra awareness for someone in the space who just stumbled across this article.

The Maverick Hackathon

I mentioned taking part in a Hackathon earlier. It was organized by Maverick Research, and spoiler alert, I did not win the money (I hear I placed third though). The concept was barcodes, OCR processing, and visual identification: I learnt a lot about barcode processing in Python (using Pyzbar), and I experimented with open multimodal models on HuggingFace, in this case Z.ai’s GLM-4.6V-Flash and Alibaba’s Qwen3-VL-235B-A22B-Instruct, with Cohere’s Command A Vision thrown in because I have some credits from long ago. Maybe I would have won if I had gone the GPT-5 or Claude route, but the open models really pulled their weight, and that really excites me!

Details and link to the live demo here.

Looking Forward

The IndabaX Ghana has grown beautifully. I sat in a workshop where facilitators, Chelsea and Joshua, taught participants how to build their first agents. I’ve done this before, but it was nice to reinforce the main ideas and see alternative routes to the smae goal.

I also enjoyed the Data Storytelling Expo. I think it’s something we might be interested in participating in at IPA in future events, giving the many years we have spent using data to create more evidence and less poverty.

If there’s one thing I felt was missing, it would be a focus on measuring what works. The culture of building evaluations so we can quantifiably assert that method A is better at method B for solving problem X. I didn’t see much of that, but given that I wasn’t in every room all of the time, let’s just assume I missed those discussions. Whatever the case, it was great to learn, reconnect, and engage with friends, old and new alike.

So until the next one… and the next one happens to be PyCon Ghana.

Stay safe, and burn a few tokens building something new.

✌🏾😎

PS. On my way back to Accra, I found ayigbe biscuit to buy 😃.
Nostalgia want finish me 🥹