>_ Data Science & Agentic AI
Ben Robinson
Product data science, experimentation, and causal inference.
Production ML and agentic AI at Google. Evidence that drives product decisions.
>_ Data Science & Agentic AI
Product data science, experimentation, and causal inference.
Production ML and agentic AI at Google. Evidence that drives product decisions.
I work in applied ML at Google, with a background in platform security and abuse prevention and a current focus on product data science: how users adopt AI, what experiments show, and how measurement should inform roadmap and design. My work spans causal inference, A/B testing, production ML for growth, and agentic systems that fit real workflows, with clear standards for how we model, evaluate, and ship.
I stay close to the work: code, evaluation frameworks, and modeling decisions. I care about understanding how people integrate AI into their workflows, not just whether a model scores well offline. I introduced experimentation frameworks that became the standard for how we measure intervention impact. The goal is the right product decision, grounded in evidence. My working thesis is that models will commoditize and the best integrations will win, which is why I care as much about product judgment and roadmap as I do about the model itself.
I've worked across adtech, healthcare, media, and big tech. The constant: rigorous analysis, production-grade ML, and turning data into decisions Product and Engineering can act on.
Product-oriented data science: causal inference, experimentation, and measurement so product and engineering can make grounded decisions. The discipline is in how we define the question, choose the method, and interpret the result for the user and the business.
Designing and shipping AI agents that run in production, with clear evaluation and guardrails. I focus on how users actually integrate AI into their workflows: what they trust, where they intervene, and what experiments and telemetry say about adoption and outcomes.
Partnering with Product on experiments, metrics, and roadmap tradeoffs on high-volume platforms. Classification and graph-based models where they serve user and business outcomes, plus the measurement discipline to know whether a change worked.
Modeling standards, evaluation frameworks, and review practices for production systems. How we define metrics, run experiments, monitor drift, and document decisions so models stay trustworthy in production.
ML-driven customer classification to identify creditworthy GCP customers. Shortened the identification window from 60 days to 3, with clear evaluation and alignment with Finance and Risk. Example of data science rigor applied to a high-stakes business decision.
Built and shipped an AI agent that automates quota allocation decisions, replacing a manual, vendor-dependent process. Defined evaluation and guardrails before rollout; eliminated vendor OpEx and reduced on-call burden by 90%. Rigorous measurement before scale was key.
Partnered with Product to redesign quota tier structures using ML-driven controls and experimentation. Improved platform outcomes (85% reduction in unwanted activity) without impacting legitimate users. Required tight coordination across modeling, policy, and product on what “right” looks like for customers.
Executed five data science projects delivering $9.5M in annualized medical claim cost reductions for Medicare. Production ML pipelines and causal inference within strict compliance. Owned technical direction across four data scientists; presented results to the Chief Actuary and executive stakeholders.
Product data science across industries. The through-line: experimentation, causal inference, and decisions that stick in production.
Own production ML systems and roadmap across customer growth, revenue, and access on Google Cloud. Partner with Product and Engineering on experiments, metrics, and how users adopt AI in their workflows. Background includes platform security and abuse prevention; current emphasis is product data science and evidence-based roadmap decisions.
ML and experimentation for GCP products; partnered with Product and Engineering on controls, metrics, and intervention design at scale. Introduced causal inference and experimentation as the standard for measuring what works.
Aetna (CVS Health)
Production ML and causal inference for Aetna Medicare, within strict compliance. Owned technical direction for four data scientists; rigorous evaluation and clear reporting to executive stakeholders.
Dotdash
Content recommendation and monetization: seasonality and text-similarity modeling at scale. Cross-functional work with Product and Ad Ops during a volatile period; focus on stable, interpretable results.
AppNexus
Revenue prediction and distributed pipelines for real-time bidding and programmatic advertising. Data science in a high-throughput, production environment.
City College of New York
Washington University in St. Louis
Collaboration, questions, or a conversation about product data science, experimentation, causal inference, or AI in production. Please reach out.