A fourth-year PhD with no top ML conference papers is not a failure. It is a signal, and the sooner you read it, the more options you have. The question you are wrestling with, whether to chase publications or pivot toward engineering, assumes those are the only two paths. They are not. You are really asking whether your time is better spent playing a game you are losing or building leverage you can actually use.
The pressure to publish at top venues is real, but it is also a legacy metric. The industry is shifting faster than the review cycles that gatekeep those conferences. Look at what is happening in applied AI: Hermes Agent developer secures $90M to bring AI agents to business users shows that capital is flowing to people who can ship working systems, not just accepted papers. Meanwhile, From ad spend to asset creation, Melius raises $20M to reimagine marketing tools demonstrates that the real bottleneck in AI adoption is not research breakthroughs but practical integration into workflows. That is where engineering skill, deployment experience, and product sense become more valuable than another citation.
Your PhD has already given you something most engineers never get: the ability to frame problems rigorously and evaluate solutions critically. That is not wasted if you pivot. What you need now is a portfolio that proves you can build. The research community increasingly rewards this too. As our Beyond the Search Box: Defining Real Research in AutoResearch piece notes, real research is increasingly defined by systems that work, not just ideas that sound good on paper. A deployed tool, an open-source library, or a well-documented engineering project carries weight that a rejected paper draft cannot.
The practical move is not to abandon your PhD but to redefine what you optimize for. Use your remaining two years to build something visible. Pick a problem that interests you, implement a solution end-to-end, and put it where people can see it. If a publication comes out of it, excellent. If not, you walk into the job market with demonstrable skills and a concrete story about what you can do. That story is worth more than a third-round resubmission. The question is not whether you are good enough for top venues. It is whether you are willing to stop measuring yourself by them.