November 27, 2019

Finished & deployed our MVP!

Burak Karakan @karakanb

We have been working hard for the last 4 weeks and managed to get our MVP live with real data. We have mostly worked on weekends and nights, got together as much as we can, split the work and we have successfully deployed the first live functional version last night.

Here it goes:

The whole process was an incredible learning opportunity for us, starting from idea to validation to implementation. We have collaborated with many tipsters through Twitter, gathered their feedback on our initial idea, did research about who is using what to gather information regarding games and bets, and we have came up with the feature list of MVP.

One of the hardest parts was to have granular tasks that could be parallelized among us so that no one has to wait for the others. To implement that, we have created separate tickets for each feature we need for the MVP, as well as tasks related to marketing, feature/idea validation, and communication. After we have created a high-level overview of what needs to be done, we started creating subtasks for each discipline, which allowed us to see all the concrete steps we need to take.

There are two Burak's in the team, and both of them are software engineers, and they have took the main responsibility of implementing the user interface as well as the backend functionality to get the product working, and throughout the process, Batikan, a software engineering student, took parts of the website to implement the full functionality. Cem and Sabri are both data scientists, and they have split the load of implementing a service for retrieving, analyzing and prioritizing the data, as well as implementing the initial artificial intelligence algorithms for smart game suggestions.

From the beginning of the project we have decided to spend our time wisely, which meant using the tools we already know for the whole process., as well as not spending our precious time on repetitive tasks. We have had previous projects with the same people, and we already had a tech stack in mind. We have used Laravel as our main backend, with Vue for frontend and Tailwind for CSS. For the AI and predictions, we have used Python with FastAPI in order to serve the suggestions to our main backend. From the infrastructure perspective, we have used Docker images for everything in order to make things as smooth as possible for all of us in the development process, and docker-compose was the main tool for gluing all the services together within this setup. When it came to deployment, we have used DigitalOcean's Managed Kubernetes as it is very easy to operate, basically nothing to do from our perspective, and we have already had Kubernetes & Helm configuration for a very similar setup before, and we already had some other projects running on the same cluster as well, which meant we got up and going with Kubernetes in around 3-5 hours. Also, since we are heavy users of GitLab, we have utilized their amazing CI feature to build ourselves a pipeline that would run the tests, build the images and deploy them to our Kubernetes cluster for every push to our master branch.

Overall, we are quite happy with the result as we have our functional MVP that works as expected, we have a good base that we can build upon with feedback from users and we have managed to achieve all these in 4 weeks, mainly weekends. It has been quite a journey up until now, and we are incredibly excited to see what future will bring us with TipMarkt.

Loading comments...