Projects

Exercism App for macOS Link to heading

• Exercism is an online, open-source, free coding platform that offers code practice and mentorship on programming languages. Since the platform can only be used online we(I and Dela) decided to build the macOS app for it.
• Our goal was; to have fun and learn a couple of new things.
• The journey
  • The first unforeseen challenge was for me was that I hadn’t built an app for macOS before. Although we would be venturing in territory I have played in before(apple), I was surprised at how little learning material there is on macOS development.
  • We started off by learning SwiftUI, something both of us hadn’t used extensively. It was fun but also really hair pulling. SwiftUI is a very declarative language and coming from a UIKit background things felt tough. Also, what’s up with that new Navigation API Apple?
  • Then we started to build this thing. Very quickly my UIKit brain led me down the path of MVVM+C. As we explored the architecture to use, I quickly discovered how using MVVM was overengineering. Combine is built into SwiftUI and so the views are already reactive and so there’s no need to have something else observing them. So, how do we do this better?
  • In comes Async Await, she’s the new kid on the block. She’s shiny, she’s pretty but does she live up to the hype? We decided to dip a toe in and so only the App is built using this. The Swift API is still using completion handlers.
  • Accesibility is also a big part of this process. I have been paying attention to the accesibility APIs that apple provides and so while building this, it was also a large part of our learning.
  • Design, this has been fun for me especially and I am still learning. I know how to build a simple product but I struggle with making it appealing.
• The Goal

• In the next month or so I want to put this thing out in the public

Vaazi Link to heading

• Vaazi is a Kiswahili word that means outfit. This an e-commerce platfrom I am building with a bunch of insanely talented individuals.

• The biggest component of this is image search. We are building a searching feature using computer vision.

• The challenges

  • The team has very limited machine learning knowledge but very big aspirations
  • Computer vision is a tough area, Fashion machine learning on the other hang is tougher than we anticipated

• All the fun

  • There is a lot of courses on machine learning available and I have found a lot that have made things easier
  • Pinterest have extensively documented their process of building the image search feature and we have learnt so much from them
  • MMFashion has also been ver helpful and we are using the open source models to customise them for our usecase

• Once a Swift girl, always a swift girl? Nope!

  • The other component of this is the front end. Since the backend has been so tough to build we wanted to get an easy front end and Flutter gave us exactly that.
  • The journey so far is filled with a lot of learning and staring deeply at pull requests. Flutter reminds me a lot of SwiftUI and I am enjoying the familiarity.

• The Goal

• By the end of the year we want to have pre-trained models that we can use. Then we will embark on the goal of fine tuning the models for our usecase. There will also be a lot of testing
• Keep working on the front end