DevFest Mbale 2018
I just spent four nights in Uganda. Here is a summary of what happened.
I attended the Sub-Saharan Africa Summit organized for Google Developer Groups (SSA GDG Summit). I applied to speak at the event, and my application was accepted. I did a presentation on using Keras for Machine Learning. Keras is a high-level interface for TensorFlow.
After my session, someone named Nsubuga Hassan approached me, introduced himself, and went on to tell me about the work he was doing with image detection on Android. He then asked me to attend his DevFest, which was coming up on November the 3rd. I agreed to make the trip without even looking up the location on a map.
When it was time to book my trip, I decided to make use of the GDE Travel Policy, which provides me with regional travel and accommodation. I went on the platform and put in my destination, and nothing came up!
Well, if Mbale isn’t there, what about Kampala? Nothing! The problem is that if I can’t find a flight to a city, I can’t book accommodation either. I resorted to Google Maps to solve the mystery.
It turns out there is only one airport in Uganda, located in Entebbe. I had to contact Hassan to explain the predicament. I could get a flight to Entebbe and back, but his team would need to take me from Entebbe to Mbale and back, and also accommodate me. He was fine with such an arrangement, so I made it.
At this point, he also mentioned that he had set up a session for the teaching staff of the universities in the vicinity so they could learn about Machine Learning. This was fine with me. I saw on Google Maps that the trip from Entebbe to Mbale was upwards of six hours, so I had to arrive at Entebbe the day prior to the first session.
I arrived Entebbe on the 1st of November at a little after 1am, and checked into a hotel that was reserved using the travel policy. This would let me rest a bit and grab a much needed breakfast before the road trip.
Hassan showed up at 10.30 and we set off for Mbale.
“You are going to see Uganda through the eyes of a local”, he said to me. I am grateful that I did.
We took a taxi, and I immediately started understanding Uganda. There are many dialects, there is English, and then there is Ugandan English. This is enough to get us to put on our thinking hats if we are going to have language translation for Uganda.
I observed a lot of things. What do they cultivate, process, and eat. What sorts of mobile devices do people use? How much processing power and memory do they possess? I hope you get where all this is going?
We got in sometime after 5pm, and Hassan left me to freshen up before dinner. I checked into Mbale Travellers’ Inn, and that was my home for the next three nights.
November 2nd was the session with the teaching staff. I asked about their level of exposure to Machine Learning, and it was decided that I should do an introduction.
I opted to explain Machine Learning from the perspective of what tools Google has made available for interested persons. I started with ML for application developers and discussed Dialogflow, Firebase ML-Kit, and Cloud ML APIs. I did a brief demonstration of Dialogflow to drive home the point.
I then went on to explain AutoML and where it comes in. I showed the interface, but did not have the right training data (nor the time) to train a model. I hope to get them something to see soon.
I completed the presentation by explaining TensorFlow and its high-level APIs, Keras and Estimators.
The feedback was generally positive, and we will need to do something to get the attendees to start teaching their students to use these APIs and tools. My slides for the session are available here.
November 3rd was DevFest Mbale. Months of work by the organizers finally came to fruition. I was the first speaker immediately after the keynote address. It was decided that the audience was new to ML and would need an extremely gentle introduction.
I decided to keep the session strictly to ML-Kit, Dialogflow, and Cloud ML APIs. I was able to successfully do this, but during the Q and A, someone asked a question that caused me to introduce AutoML.
At the end, the attendees were interested in the kinds of applications they could build, but were also interested in solutions that would cater to their accents, and their native dialects. I told them they would need to get into research to see these solutions built.
We distributed t-shirts to attendees by making them answer questions. I asked 15 questions, and they were all eagerly answered. The audience actually paid attention.
The 4th saw us leave Mbale for Entebbe so I could catch my flight. This trip revealed a few important things:
- Having foreign guests reinforces the importance of the technologies we are asking people to master, as well as a career path. A lot of the people that would not have turned up did so because of the presence of foreign speakers.
- You can’t solve a problem from a distance. You can’t even detect it without being there in person.
- People see the problems around them, they just aren’t aware that they can start building solutions to those problems, or go into research to solve those problems. I look forward to some of these attendees applying to either AMMI or Google Brain in Ghana so they can work in the NLP domain.
- Travel on the continent is hard, but we will keep doing our best.
I am grateful to the Experts Relations team for this opportunity to be at Mbale.
I am putting together an album of some photos I took. You can view it here.
