open-life-science · yochannah · Sep 3, 2025
diff --git a/_posts/2025-09-03-nyasita-third-indaba.md b/_posts/2025-09-03-nyasita-third-indaba.md
@@ -0,0 +1,56 @@
+---
+layout: post
+title: "My Third Deep Learning Indaba Experience"
+authors:
+- nyasita
+image: https://images.unsplash.com/photo-1633158829875-e5316a358c6f
+photos:
+  name: Towfiqu barbhuiya
+  license: CC-BY
+  url: https://unsplash.com/@towfiqu999999
+---
+
+
+This year marked my third Deep Learning Indaba. When I sent in my application to attend, I remember being anxious, wondering if I would be selected. I felt that way because my first application back in 2023 felt more like a try-out. I didn’t even expect to be selected, but I was, and I even ended up presenting a poster based on part of my master’s research, where I used AlphaFold to predict drug targets. The second time, in 2024, I was part of the organising committee in the programs department, so I didn’t have to submit an application at all. But this year I had to apply again, this time as an attendee without work to present, just with the intention of networking, enjoying the event, and being a sponge; to soak in knowledge. For the past two years, the Deep Learning Indaba has been happening in West Africa; in Ghana in 2023, then in Senegal in 2024. So when it was announced that in 2025 it would be hosted in Kigali, Rwanda, in East Africa, I was stoked. Even though I’d never been to Kigali before, it already felt close to home: just a short flight (or even a bus ride) away. 
+
+The [Deep Learning Indaba](https://deeplearningindaba.com/) (DLI) (or simply the Indaba for me) is the annual meeting of the African machine learning and AI community. Since its start in 2017, the [mission](https://deeplearningindaba.com/about/our-mission/) has been to strengthen African AI and ensure Africans are not just consumers but owners and shapers of the field’s advances. This year’s theme *Urunana*: Hand in Hand for AI in Africa, perfectly captured the spirit of solidarity and collaboration that defines the Indaba.
+
+![][image1]
+
+*(An image of delegates at the Deep Learning Indaba 2025 at the University of Rwanda in Kigali, Rwanda. PC: [The Deep Learning Indaba](https://www.linkedin.com/company/deep-learning-indaba/posts/?feedView=images))*
+
+There are usually three things I look forward to at the Indaba… wait, no, four:
+
+1. The people. When I interviewed [**Siobhan Mackenzie Hall**](https://we-are-ols.org/people.html#smhall97) (an OLS mentor and one of the OGs of the Indaba) back in 2023 for my podcast *The* *Science in Real Life*, she told me: *“Look forward to the people.”* She was right. The Indaba community is warm, curious, and adventurous. Exactly my kind of people. This year, I reconnected with Siobhan, and I finally met [**Ahmed Unshur**](https://www.linkedin.com/in/ahmedunshur/) and [**Tajuddeen Gwadabe**](https://we-are-ols.org/people.html#tajuddeen1) in person. I’d seen their names in OLS co-workings and other meetings, but we hadn’t crossed paths face-to-face until now. It was refreshing to chat about our work and share highlights from the Indaba. They’re doing amazing things, by the way!
+
+| ![][image2](Nyasita and Tajuddeen Gwadabe at the Deep Learning Indaba 2025 in Kigali) | ![][image3](Nyasita and Siobhan Mackenzie Hall at the Deep Learning Indaba 2025 in Kigali) | ![][image4] (Nyasita and Ahmed Unshur at the Deep Learning Indaba 2025 in Kigali) |
+| :---- | :---- | :---- |
+
+2. The tutorials. My background isn’t in AI or coding per se, so I find the tutorials incredibly valuable. They’re like friendly lecture-style intros that ease you into the technical depth. My favorite this year was from Luis Serrano, who explained something that completely reframed how I think about ChatGPT. He said, *“The thing that drives me crazy about ChatGPT is that it generates one word at a time.”* At first, I was baffled, because as a user, it feels like it writes whole paragraphs in seconds. But from what I understood, what he meant is that ChatGPT is a predictive model: at every step, it selects the most likely *next word* based on everything it has generated before. What makes it powerful (and better than phone autopredict) is its ability to look far back into the context, sometimes thousands of words, so the sentences come out coherent, not just like WhatsApp or message predict suggestions strung together. That clicked for me.
+
+3. The practicals. The Indaba wouldn’t be the Indaba without the hands-on sessions in Colab notebooks. This year, I stuck to the LLM track and loved the “behind-the-scenes” look at generative AI. Two stood out:
+
+   * Exploring the World of Generative AI – We used Stable Diffusion to remix sketches and photos, played “AI or Real?” guessing games, and discussed ethical tensions around bias, deepfakes, environmental cost, ownership, and global equity. AI has become so creative and accurate that it’s hard for us to tell AI-generated images from actual images. Like, look at these two below, are you able to tell which one is real and which one is AI? 
+
+| ![][image5] | ![][image6] |
+| :---- | :---- |
+
+(The pizza is AI, and the mountains are real. I know, shocked me too!)
+
+* Large Language Models under the Hood – We unpacked Transformers, attention, tokenization, and embeddings. Beyond using LLMs like ChatGPT or Gemini, we even trained a small Shakespeare-inspired model ourselves. These are things I didn’t learn in school, and thanks to the Indaba, I am learning them. 
+
+4. The culture dinner. Nothing screams *ubuntu or umoja* more than this night of African dress, music, food, and celebration. Even after three Indabas, I’m still as excited as I was at my first culture dinner. But unlike that first indaba, I tried to represent Kenya as much as I could, adorning Maasai accessories with my somewhat shein-looking outfit. *(for image reference, see the picture above with Tajuddeen)*
+
+*![][image7]*
+
+*(Image of DLI delegates at the DLI culture dinner at the University of Rwanda in Kigali. PC: [The Deep Learning Indaba](https://www.linkedin.com/company/deep-learning-indaba/posts/?feedView=images))*
+
+Beyond the Indaba, Kigali itself was an adventure. When I arrived, I couldn’t help but notice the sheer number of *motos* (motorbike taxis). We have plenty in Nairobi, but Kigali takes it to another level. Here, *motos* are the city’s main mode of transport, weaving through the hills just as fast as any car could. People also walk, but if you decide to walk, be ready: every stroll feels like a mini hike because of Kigali’s hilly landscape (my legs can confirm this). Kigali is indeed clean. If I did [the white-socks test](https://www.tiktok.com/@jimbo.h/video/7480630850036567298), I would easily pass here! (Yes, the double entendre is intended.) I also quickly learned that while Rwanda is multilingual (English, French, Swahili, and Kinyarwanda are spoken there), most people prefer Kinyarwanda for their day-to-day communication, and thanks to this, I have added four new words to my vocabulary: *urakoze* (thanks), *muraho* (hello), *murakaza neza* (welcome), and *urunana* (hand-in-hand). 
+
+I also made time to explore different places in Kigali, starting with the [Kigali Genocide Memorial](https://kgm.rw/), a deeply moving space that preserves the memory of the 1994 genocide in Rwanda. It was a sobering reminder of resilience and the importance of building communities rooted in care and solidarity; something the Indaba itself embodies. On a lighter note, I visited [Nyandungu Eco-Park](https://www.nyandunguecopark.rw/aboutus) and discovered [Fazenda Sengha](https://sengha.com/), a hidden gem in Nyamirambo district. Between quad biking, zip-lining, and archery (where I hit a bull’s eye on my very first attempt!), it was the perfect way to balance the intensity of the Indaba with a little adrenaline and joy. Oh, and I also managed to fall into the swamp at Nyandungu—unforgettable, I know!
+
+This Indaba gave me what it always has: new knowledge, new friendships, and a deeper sense of belonging in the African AI community, but this time with Kigali’s breathtaking hills and vibrant spirit as the backdrop. Next year, the Deep Learning Indaba heads to **Nigeria**, and I’m stoked! I must finally settle for myself which jollof rice reigns supreme, Ghanaian or Nigerian. I’ve already had Ghanaian, but right now for me, the Senegalese jollof is number one.
+
+If you’re an OLS member and want to come to the Indaba, join [their newsletter](http://eepurl.com/ii4Llb) to stay up to date, especially for when applications open. And if you’re selected, be sure to connect with other OLS-sers, they will surely be there (and maybe me too). See ya!
+
+*Urunana*, let’s keep advancing AI in Africa