Machine Learning: Proactivity and Key Planning

Proactivity, Blocking Time, and Planning

As we progress in our careers, whether over a few months or several decades, certain elements will remain constant in our professional environment. Projects, for example, are organized initiatives that drive us toward specific goals. This month, I gained valuable insights into managing these projects in the field of machine learning.

Being Proactive Ensures Smooth Progress

In our daily work, we often face projects that we dread, but also those we enjoy and wish to spend more time on. Whether we like them or not, these projects typically span long periods. They are not ends in themselves, even though they can sometimes give that frustrating impression. In reality, they are designed to bring us, or our company, closer to a defined goal.

In the context of machine learning, these goals can vary: delivering a model to a client, writing a scientific paper, or setting up an MLOps pipeline. All these projects require sustained attention over an extended period and, most importantly, the support of others.

This support does not necessarily mean that others must actively push the project forward, although that is always appreciated. Rather, it involves receiving the necessary resources to make progress. Sometimes, it can be as simple as obtaining permission to use a specific computing resource. Other times, it may involve more significant decisions, such as approving the purchase of essential software.

Projects rarely proceed smoothly, with a constant tailwind. On the contrary, each step can become a potential obstacle. What I have learned is that proactivity can prevent many problems from arising in the first place. Developing this skill is essential, not only for machine learning projects but also for autonomy in general: the ability to direct one’s actions deliberately and seek solutions independently.

In the context of ML projects, proactivity can manifest in several ways: requesting approvals in advance, developing contingency plans, anticipating alternatives, or allocating more time from the outset to create a buffer.

Blocking Time to Accomplish Projects

After emphasizing the importance of proactivity in avoiding obstacles, another lesson emerges: to make progress, one must also be proactive in blocking time for their projects.

This may seem obvious, like many important things once understood. However, the fact that something is obvious does not mean it is easy to implement.

Let’s consider the typical day of a machine learning practitioner. It doesn’t matter whether they are in research, engineering, or administration; the difference lies in the projects they are working on.

But here’s the catch: it is almost never just one project. Often, there are multiple projects underway.

Our ML practitioner likely has more than one project on their desk. There is the main project, such as creating an MLOps pipeline, writing a paper, or upgrading a computing cluster. And then, as any PhD student can attest, there are the other "secondary" projects: presenting results, teaching courses, daily administrative tasks. All these elements require time and attention. And we return to the main project: the time spent on other projects is not available for the main project.

So, how can one dedicate more time to the main project without neglecting the others? The answer is quite simple: block time in your calendar.

Any free slot in your schedule can prompt others to invite you to a meeting. By simply blocking out time slots, you can ensure that you dedicate enough time to your main project. Unblocked time remains available for other projects.

Ultimately, it comes down to prioritization in 90% of cases: prioritize the main project. In the remaining 10%, emergencies may require deviating from this rule.

Planning, Planning, and Sticking to the Plan

Reflecting on the past month and the two previous lessons, a general conclusion emerges: planning is crucial. And: it is essential to maintain the plan as it is.

In our ever-evolving world, there is always something new. Take, for example, the notebook with which I am writing these lines; it dates back to 2020. Since then, five new versions have been released.

Or consider: do you remember GPT-3? We are now at GPT-5.4, and ChatGPT has become multimodal.

Or, if more examples are needed: the news. Every day, there is something new. All this to say: if you plan something, it is easy to set the plan aside to do something else instead.

This might be acceptable, but excelling in a field requires that we dedicate time to it again and again. This essentially means being proactive, blocking time, and… planning. Whether by literally writing a plan or forming it semi-consciously in your mind.

For the ML projects we have discussed here, nothing would be accomplished without planning. Neither the article, nor the new material, nor the pipeline.

If you plan well enough, without being overly precise, you can achieve great things. But only if you ensure that the plan remains the plan, undisturbed by the latest news.