Staff Machine Learning Engineer

Job Description About the team At Zendesk, we truly believe that to build great products you have to have great people. We enjoy working with other smart, focused people who care about both the products and the code they write. We value collaboration and release frequently. We like and use agile processes and believe that pragmatism always triumphs over dogmatism. We all own the product or service we work on and enjoy the impact we have improving it. Our mission is to elevate Zendesk’s routing and presence products to their next stage — from a rules-based engine to an agentic routing engine where intelligent, adaptive decisioning is the core differentiator. Routing in a realtime, omnichannel world is a complex problem domain that needs a robust, scalable and maintainable solution. Our recent partnership with the Machine Learning team delivered Predictive Routing to GA; this role exists because the next phase of the roadmap needs that capability to live inside the team. About the role We’re hiring a Staff Machine Learning Engineer to be the embedded ML expert for Routing & Presence. You’ll own the ML surface of our routing products end to end — from feature engineering and model design, through experimentation, to production serving, monitoring and continuous learning. You’ll work side-by-side with backend engineers, product managers, and the central ML platform team, and you’ll set the technical direction for how ML shapes our agentic routing engine over the next few years. This is not a pure research role. The bar is: can you take applied ML knowledge and turn it into measurable customer outcomes, reliably, at scale, inside a product engineering team? What you’ll do: Own the models and algorithms at the heart of Routing & Presence. You are accountable for their quality, reliability, and interoperability with the rest of the business — from Predictive Routing today through the agentic routing engine we’re building next. Plan and scope ML work with observability and iterability designed in from the start, and with a clear view of how the resulting algorithms overlap and interoperate with adjacent systems. Propose and evaluate alternatives. For any meaningful design decision, surface the candidate approaches, the tradeoffs, and the reasoning — don’t jump to the first plausible solution. Design experimentation frameworks (offline evals and online A/B) tailored to routing outcomes, with statistical rigour and a clear tie-back to customer-facing metrics. Lead innovation sessions with Product ahead of the product development cycle — help shape what we build, not just how . Identify new data sources needed for learning and evaluation, working with Product and Data to bring them into the pipeline responsibly. Combine multiple signals — ticket sentiment, agent skill, historical performance, realtime load — into dynamic routing decisions, with explainability and guardrails admins can trust and tune. Shape the boundary between classical ML, LLM-driven components, and rules ; decide when each belongs where, and why. Anticipate issues across the full lifecycle — development, testing, deployment, operations, and support — and build mitigations in, even where you aren’t the specialist. Work across multiple teams and systems with PMs, backend engineers, the central ML team, and adjacent product teams to define problems and set the high-level direction of solutions. Mentor scientists and engineers — raise the team’s ML literacy, support teammates through their own challenges, and make sure learning opportunities are distributed fairly. Look ahead 18 months. Make sure what we design and build today still serves us as the agentic routing engine matures and the customer base grows. Build reputation as a subject-matter expert — share best practices with the wider science and engineering community at Zendesk and beyond (writeups, talks, internal forums) where it’s useful to do so. What we’d like from you You are an open, thoug ... (truncated, view full listing at source)