The thing I really want to avoid is specializing myself into one company, getting really out of shape, and then being let go and realizing I no longer know how things work and have to start over with my career. The tech sector moves fast, and layoffs have been happening for a while. How do I inoculate myself against that? Or at least minimize the impact and make it easier to bounce back?
Where to Go
First off, staying sharp in any market depends on where you want to go, and sometimes the only way to figure that out is to get lost. Getting lost can build confidence, similar to driving without a GPS to learn an area. For a while, I thought I wanted to be a manager, partly to be a leader and mentor. Over time I realized those things aren’t restricted to managers. In fact, they can be harder to do as a manager since so much time goes into managing the team. If I hadn’t been unsure about whether I wanted to manage, I might not have figured out what I actually wanted to do.
As a data scientist, there are a few different paths after you’ve broken into the field and started to level out. You can move into management or a lead role and manage a team. You can switch functional IC roles like data engineering, machine learning, product or project management, each with its own ladder. You can move into a new domain like risk, marketing, or finance, or switch companies just to change environments. You can stay put and focus on work life balance, spend time on things outside of work, or progress as an IC by taking on larger scope projects like migrations or cross functional work. You can further specialize in one area or expand into side projects like mentoring, consulting, writing, or entrepreneurship. I’ll call another category staying sharp.
By the end of that list, some of these directions blur together. Managers can still work on large projects. Switching roles or companies can mean taking on more scope. These paths aren’t mutually exclusive, and priorities can shift quickly as life changes or new opportunities appear. As layoffs become more common in tech, I’ve become more aware of the value of having a data science portfolio that lives outside any one company (e.g. my github and this blog!). Having something in the cloud that you can take with you regardless of role has started to feel increasingly important, and it’s something I’ll expand on later.
Tools and Skillsets
In terms of staying sharp in data science, toolsets can change dramatically. Vertica becomes Snowflake. Excel VBA becomes Airflow/DBT. SAS, Matlab, or SPSS becomes Python or R. Tableau becomes Looker, Amplitude or Logrocket, though there’s still plenty of SQL underneath. Googling becomes ChatGPT or Claude Code. Each of these tools is a business, and platforms shift as contracts end, companies outgrow solutions, or industries move on. Some skills transfer cleanly, others don’t. It can be surprisingly painful to lose a UI you’ve used for years and realize how many shortcuts you relied on without noticing.
Even ideas about skill sets change. Generalists versus specialists. T-shaped people. PaintDrip people. The hard truth is that even if you do everything right, you might learn a skill that doesn’t matter in five or ten years. I specialized in FAME and RATS at the Federal Reserve Board. Have you even heard of them? Nothing lasts forever, and not everything transfers directly. What mattered more was what sat underneath. I learned how to code, debug, and run regressions. RATS stands for Regression Analysis Time Series and FAME stands for Forecasting And Modeling Environment. With the full names, they sound a lot closer to modern data science, and much of what I learned during those years transferred, even if the language changed.
How to Stay Sharp
I think of staying sharp as staying marketable as a data scientist. Avoiding a situation where your skill set becomes obsolete and you can’t find a job no matter how hard you try. Doomscrolling LinkedIn doesn’t help. What does help is interviewing. I know that sounds obvious, like asking how to stay in shape and being told to exercise, but hear me out.
Companies have work that needs doing, so they make decisions about headcount and write job descriptions to fill those roles. They’re willing to pay for it because the work matters to them. If you see a job description, someone wants something important done. Those descriptions include tools like SQL, Python, or ETL systems, but the tool is just the solution to a problem. The more important part is understanding the problem itself. In that sense, this is similar to how product managers think. The data scientist is the product, the skill set is the feature set, and the customer need is the job description.
If you collect enough of those signals, you get a sense of market demand. Job descriptions aren’t always accurate representations of the work, so interviewing helps fill in the gaps. You learn what companies actually want and how your skills line up, or what you might need to learn next time. Interviewing isn’t a commitment. It’s just expressing interest and following through on it. Sometimes you even realize you genuinely want the role. Nothing is real until there’s an offer, and you’re not taking someone else’s place. If you leave your role, that seat opens back up.
You can also gather information closer to home by looking at your own company, a friend’s company, or your broader network. Interviewing can even help you notice valuable problems internally that you might have missed. Knowing how to spot meaningful problems is core to being a data scientist. At the end of the day, we’re problem solvers trying to understand what’s going on with customers.
Before careers were a thing, people just had jobs. They found work that needed doing, did it, and went home. The principle hasn’t changed, but stability can blur it. Over time, it’s easy to cling to the status quo instead of staying connected to the actual work you’re paid to do. Stability isn’t bad, especially in late stage capitalism, but sometimes foundations shift for reasons outside your control.
You don’t need to interview constantly. I usually respond to a recruiter once a month or once a quarter, even if it’s just a short conversation. It helps me learn what’s out there and see if there are better solutions to problems I’m already thinking about. I also try not to interview during the first year of a role unless I really hate it, just to give it my full attention. This doesn’t need to be frequent. It just shouldn’t be something you think about once every five or ten years.
Core Transferable Skills
Some skills transfer across almost any role:
- Communication, being able to tell a clear story to influence and understand others (to improve writing, reading helps).
- Management, including managing your own time, energy, and identity outside of work (See my soft skills debugging article).
- Analysis, being able to read a situation, develop a strategy, and execute on that insight. Said simply: keep your BS detector strong and keep learning.
- Networking, knowing real people and building trust over time. Real people sit behind every decision, even the ones mediated by algorithms.
What you don’t work on intentionally doesn’t grow. Staying sharp requires making it a priority and having a plan, even a loose one. Sometimes you get out of shape and get back into shape. That’s fine. Life is messy. I also find it helpful to zoom out beyond data science and even beyond the business itself. How is the industry doing? Does the company create real value for its customers? Connecting those bigger questions back to your work helps keep things grounded. Companies need data to avoid flying blind. They need experiments and insights to learn and make decisions. Someone has to do that work. It might as well be you.
Daeus Jorento 