Product Manager, AI/ML - Agent Platform (ML Models, Evals & ROI)

ZoomInfo (ZI) is how businesses go to market (GTM), and we're on a mission to modernize go-to-market for everyone. Powered by real-time data and insights, our unified GTM intelligence platform helps sales and marketing teams find, acquire, and grow customers. Within ZoomInfo, the Agent Platform team is building the execution layer for the GTM AI OS - Agent Teams, Agent Studio, and the underlying agentic infrastructure that is reshaping how thousands of revenue teams operate every day.

Agents are only as smart as the intelligence layer that feeds them. This role owns that layer. You will lead product for the portfolio of ML models that power ZoomInfo - the scoring, ranking, recommendation, classification, and signal models that decide which accounts to prioritize, which contacts to surface, which signals to act on, and which plays to run. You will own the evaluation frameworks that prove those models are working, and the ROI frameworks that prove the agents built on top of them are delivering real business value to customers.

The data you work with is ZoomInfo's unfair advantage - the deepest B2B contact graph in the industry, Chorus conversation intelligence, intent and scoop signals, firmographics, and behavioral data — and the models you ship will turn that raw data into the decisions hundreds of thousands of sales and marketing professionals trust to run their day.

What You'll Do

Own the ML model portfolio. Lead product strategy for the suite of ML models powering Agent Teams — account/lead scoring, ranking, recommendation, propensity, classification, intent prediction, next-best-action. Set the roadmap for which models we build, which we improve, and which we retire. Decide what data each model needs, what business decision it serves, and what good looks like.

Build the evaluation system. Define and operationalize how we measure ML model quality at every stage of the lifecycle. Offline metrics (precision/recall, AUC, calibration, lift, NDCG), online experimentation (A/B tests, holdouts, interleaving), production monitoring (drift, stability, fairness, segment-level performance), and the regression suites that catch degradations before customers do. Make model quality a transparent, first-class product surface.

Prove agent ROI. Own the framework that connects ML-driven agent behavior to customer business outcomes — meetings booked, pipeline generated, win rate lift, cycle time reduction, rep hours saved. Build the attribution methodology, instrument the funnel from model output through agent action to revenue, and turn that data into the proof points that drive expansion, renewal, and competitive wins.

Set vision and roadmap. Paint a compelling vision for the intelligence layer of the Agent Platform and translate it into a quarterly roadmap aligned with platform and company objectives. Sequence model, eval, and measurement investments so they compound.

Define and analyze metrics. Identify the metrics that matter — model lift, recommendation acceptance rate, agent task success rate, AI Credits Consumption per outcome, cost per successful action — and build the dashboards that make them visible. Translate insights into prioritized initiatives using Amplitude, Tableau, and your own SQL/Python.

Engage customers. Sit with customers regularly to understand where the models help, where they miss, and what would make them indispensable. Review Chorus recordings and product sessions weekly. Bring that ground truth into every model and roadmap decision.

Lead cross-functionally. Galvanize data scientists, ML engineers, software engineers, designers, and GTM partners around a shared bar for model quality and business impact. Be the focal point that turns ambiguous tradeoffs — precision vs. recall, coverage vs. quality, model complexity vs. interpretability — into clear, defensible decisions.

Drive stakeholder alignment. Communicate model strategy, eval results, and ROI through crisp written narratives — PRDs, exec updates, customer-facing materials — that align leadership and rally the broader org behind the work.

What You'll Bring

Deep ML product instinct. You've built and shipped ML-powered products at scale — scoring, ranking, recommendation, classification, or propensity systems that real users depend on. You understand the full model lifecycle: data, features, training, validation, deployment, monitoring, retraining. You have strong opinions about what makes an ML product work and you back them with data. 6+ years of product management experience, with the last 2+ focused on ML or data products.

Evals are second nature. You understand the difference between offline and online evaluation, why offline metrics can mislead, and how to design experiments that isolate model impact from confounds. You know how to think about calibration, drift, segment-level performance, and counterfactual evaluation. You've built or scaled ML eval infrastructure before, or you're hungry to.

You can quantify value. You're fluent in connecting model behavior to revenue impact. You've built ROI models, instrumented funnels, and used the resulting data to drive expansion, pricing, or roadmap decisions. You think in unit economics — lift per recommendation, value per successful action, margin per agent run.

Data is your passion. You've built complex data products before and aren't put off by personalization, ranking, or recommendation problems — in fact, you often find yourself trying to figure out how a company knew exactly what you wanted.

You thrive in ambiguity. ML product work means tradeoffs without clean answers — model accuracy vs. coverage, freshness vs. stability, complexity vs. explainability. You prefer problems with no clear answer, you go deep to become the expert, and you make confident calls when others freeze.

You write to think. You compress messy ML tradeoffs into one-page narratives an exec can act on. PRDs, strategy memos, customer-facing positioning — high bar, produced quickly.

You're hands-on with data. Amplitude and Tableau are reflexes. You write SQL and Python to answer your own questions rather than waiting on an analyst. You can read a model evaluation report, ask the right follow-up questions, and instrument what you ship.

You partner deeply with data science and engineering. DS and ML engineers see you as a trusted thought partner — someone who sharpens the problem, defines metrics carefully, and helps decide what's worth modeling. You can have a real conversation about feature design, label quality, training/serving skew, or experimentation tradeoffs.

Customer obsession is your default. You'll drop a roadmap session to jump on a Zoom with a customer who's seeing a strange recommendation. You watch Chorus calls. You read tickets. You bring that voice into every decision.

You're energized by design. Model outputs only matter if users trust them, understand them, and act on them. You partner with designers to make ML-driven recommendations legible, explainable, and actionable in the workflow.

What Does Success Look Like

Every ML model powering Agent Teams is backed by a robust eval suite, clear quality metrics, and a defensible business case. Customers can see, in their own dashboards, the measurable value the agents deliver — and that visibility drives expansion and renewal.

AI Credits Consumption grows because customers are getting more value per credit — and you can prove the unit economics of the models behind every agent we run.

Data science and engineering see you as the person who turns hard modeling and eval tradeoffs into clear decisions. You make their work compound and their impact visible.

GTM leaders and customers trust ZoomInfo as the credible source on agent ROI in the GTM category — because you built the proof.

The competitive narrative against Salesforce Agentforce, HubSpot AI, and others is anchored in measurable, defensible model quality and customer value — because you built the measurement and the proof points.

How We Work

Be Relentless. We will be relentless: when we deliver value to clients; when we compete; when we run into difficult problems. We will outpace and outcompete our competition. We are smart, clever, and resourceful in everything we do. Competing with us for the same customers will be a daunting experience.

Be Entrepreneurs. Entrepreneurs hustle, move fast, take ownership, they have autonomy, and make decisions. They are accountable, but they do not operate in fear, they aren't afraid to fail. They are resilient, resourceful, and solutions-oriented — even when conditions say otherwise. We will entrust, and expect, our leadership and our teams to operate as entrepreneurs.

Be Experts. Winning for our customers requires that we understand them deeply. Winning for our business requires us to be experts in our domain areas. Great companies solve complex and costly problems for their customers — you can't do this without being an expert in your customers' day-to-day workflow. We will demand this level of customer and business expertise from everyone at the company.

Be Innovators. Innovating is who we are, it is how we have gotten here. But not just with our product and how we define what GTM means for the future. We innovate to be leaner and more efficient in the way we operate, we will find ways to drive outcomes no one thought were possible — from marketing and lead generation to legal, finance, and procurement. Through innovation, we will set a new standard for how great companies operate with discipline.

Be a Team. We can't win if we act like a random assortment of siloed groups. We are one team, working together to win. We collaborate and lean in to help each other.

#LI-VC1

#LI-Hybrid