As AI dismantles the traditional data career ladder, one senior data scientist’s trajectory offers a model for what comes next
When Ziyu Wang first entered the data analytics profession, the career path was legible: Data Engineers built pipelines; Data Scientists built models; Product Managers defined what to build and why. The roles were distinct, the handoffs were formalized, and success meant climbing the ladder within your lane.
Wang did not stay in his lane.
Over a career spanning leading organizations in the technology and financial services sectors, Wang has operated as a Senior Data Scientist, Engineer, and Product Manager simultaneously — building production-grade data infrastructure alongside advanced analytical models, developing deep domain expertise in cybersecurity, and shipping security tools with significant organizational impact. He has designed internal training programs, been invited to speak at top universities, and become a vocal advocate for rethinking how the industry develops its talent.
Today, as artificial intelligence collapses the boundaries between data science, engineering, and product management, the career Wang built by crossing those boundaries looks less like an outlier and more like a preview.
The Blueprint, Part One: Go Deep
AI-powered tools have made it possible for product managers to write their own queries, for engineers to build data pipelines with copilot tools, and for data scientists to deploy models without waiting for engineering support. Most companies have responded to this shift by investing in AI tooling. Wang argues they are overlooking an equally critical investment.
“Almost no one is investing equally in domain training,” Wang says. A large language model can write a query in seconds — but if the person prompting it doesn’t understand the business logic or how a metric is defined, the output will be technically flawless and substantively wrong. Garbage in, garbage out.”
This is why Wang’s first principle for navigating the AI transition is counterintuitive in an era that celebrates generalists: go deeper.
“AI literacy is table stakes,” Wang says. “What AI cannot replicate is genuine domain expertise — the contextual understanding that tells you which question to ask, not just how to answer it.”
Wang’s own career validates the point. As a data scientist working behind code vulnerability detection tools and a Certified Ethical Hacker, he developed a deep understanding of how security flaws manifest in code — the patterns that make certain vulnerabilities severe, the conditions under which they become exploitable, and the gaps that existing detection methods miss. That domain expertise proved decisive: it allowed him to engineer improvements that surfaced critical risks more effectively, enabling remediation before vulnerabilities could be exploited.
“A general-purpose data scientist could have worked on the same tools and delivered competent analysis,” Wang reflects. “But without understanding the security domain — how attackers think, which vulnerabilities actually matter, what the tools are trying to catch — you’re optimizing metrics without understanding impact. The domain knowledge told us where to look.”
The Blueprint, Part Two: Go Broad
Domain depth is Wang’s moat. The breadth of capability is his leverage.
“Go deep is what AI can’t replace,” he says. “Go broad is what AI now enables you to do.”
Data professionals, Wang argues, should stop thinking of themselves as service providers to product teams and start building products themselves. They already understand the backend data architecture, the analytical frameworks, and the business logic. With AI-assisted development, they can act on that knowledge directly.
Wang’s own work on vulnerability detection tools illustrates the principle. Rather than limiting himself to the data science layer — delivering models and handing off findings — he operated across the full product lifecycle: designing the analytical methodology, engineering the underlying infrastructure, and defining the roadmap that shaped how the tools surfaced and prioritized risks. Work that would typically span separate data science, engineering, and product teams, he drove end-to-end.
“If I had only done data science — run the analysis and handed off a slide deck — it would have gone into someone’s backlog,” Wang says. “Instead, I helped shape the product. That’s what ‘go broad’ means in practice.”
Preparing the Next Generation
Wang’s influence on the profession extends beyond the organizations where he has worked. As a recurring guest speaker at the University of Pennsylvania, the University of Cincinnati, and Wake Forest University, he counsels students on building careers in a field being redefined in real time.
Do not optimize for the job market that existed when you started your program,” Wang tells them. Learn the tools, but spend equal time understanding a domain you care about. The tools will change. The domain knowledge compounds.”
It is advice grounded in a specific vision of where the profession is heading. Within three years, Wang predicts, the traditional separation between “data team” and “product team” will be meaningless at most technology companies. Professionals who have embraced both depth and breadth will hold hybrid roles — part analyst, part engineer, part product builder. Those who resisted will find their roles automated or absorbed.
It is a future Wang has been building toward for years — not by predicting it, but by living it.
“This is the biggest opportunity the data profession has ever had,” Wang says. “For the first time, one person with domain expertise and AI-assisted tooling can do what used to require a five-person team. The question is whether you seize that — or keep waiting for a ticket in the backlog.
