An abrupt transition occurred when we started working from home. It was an event unforeseen and unplanned for. As a team leader, this situation needed some thinking and planning on my part as to how we could optimally work as a Data Science team.
Let me share my experience in moving to a complete virtual environment.
Working from home has affected both people and data. It is assumed that data science practitioners tend to work in an isolated environment. We are considered as the nerds who prefer to work in isolation and avoid social interactions so that we can have a completely focused time on model development.
At first glance, the work from home environment would seem ideal for us. We have our laptops and a strong internet connection, that’s all we need to focus on our work!
Alas! We were deceived. Serious Analytics needs Serious Collaboration and this is the story of our small combat.
Re-Kindling the Team Spirit
While in workspace, we could lean on each other by venting about a bug over a cup of coffee, those interactions were now subject to a meeting invite. The celebration after successfully deploying a model to production also had to wait until the next team huddle. This could have been a simple high five in office. At work, we do find time to destress and spend more time as a team in having fun. With remote working, everything became laptop centric and the boundary of personal time and office hours faded.
As a team, we focused on finding fun time every day and without fail, talking to each other everyday. In that allocated time, no one talked about work and we started our day by simply getting to understand each other or just playing games. We just need to trust our team that they will deliver their project even if we don’t hysterically check-in with them. We followed scrum stand-ups to call out roadblocks and updates. This worked very well for us.
Pulling in the data when in an enterprise set up is much easier than at home. One of the biggest challenges is network connectivity. Sometimes we might need to deploy code developed by another team member. The presence of collaborative platforms made the code deployments easier due to uniform package versions and software setups.
While we might have more isolation time for development, we didn’t have the luxury of popping up at a teammate’s desk and take their help in decoding a bug or brainstorm a modeling idea. This slowed down the pace of progress as compared to the earlier times.
Coding needs creativity! I consider analytics as an art and creative minds don’t like following standard processes. As a result, very few of us end up using the code collaboration platforms all the time. But this scenario showed us the importance of coding platforms and unleashed a different way of collaboration via pull requests and code reviews. The standard coding practices, package developments and separating the business logic from the front end, helped remove such blockers. Making code publishing a habit became a necessity v/s only being a best practice. Identifying what software each one of us uses and getting the right version and all the dependencies became a high priority. This needed great collaboration with each other and with the IT teams.
When it comes to data scientists, it becomes quite tricky to actively reduce the technical debt and come to a place of process excellence. As leaders, we have to find a balance in maintaining the uniqueness in development style of each individual and reducing the technical debt that comes with it. Process excellence is much needed but can’t be enforced as a mandate else we lose on the creativity in each solution. We learned process rigor by understanding our current state, taking baby steps to move towards process standardization and being forced by covid to take it up as the way we work.
In the end, it’s all about learning how to structure and reproduce the creativity!