Lead Data Scientist - Data Cloud Acceleration

WHO WE ARE 

Zeta Global (NYSE: ZETA) is the AI-Powered Marketing Cloud that leverages advanced artificial intelligence (AI) and trillions of consumer signals to make it easier for marketers to acquire, grow, and retain customers more efficiently. Through the Zeta Marketing Platform (ZMP), our vision is to make sophisticated marketing simple by unifying identity, intelligence, and omnichannel activation into a single platform – powered by one of the industry’s largest proprietary databases and AI. Our enterprise customers across multiple verticals are empowered to personalize experiences with consumers at an individual level across every channel, delivering better results for marketing programs. Zeta was founded in 2007 by David A. Steinberg and John Sculley and is headquartered in New York City with offices around the world. To learn more, go to www.zetaglobal.com.

About the team 

We’re a small band of business savvy technologists who treat machine learning as a means, not an end. Our charter: find revenue shaping opportunities, ship the first working model or service in days or weeks, watch how the market reacts, then double down—or pivot—fast. Ego takes a back seat to curiosity, and “good enough for now” often beats “perfect but late.” 

What you’ll do 

• Frame & focus. Translate fuzzy growth ideas into moldeable problems, pick the metrics that matter, and design bite-sized experiments to learn quickly. 

• Build fast, in or out of the box. Finetune a foundation model when it’s the 80percent solution; spin up a from scratch architecture only when the use case truly needs it. 

• Own the full lifecycle. Prototype in notebooks, productionize via Python APIs or lightweight microservices, and wire up offline scoring, real-time inference, and monitoring. 

• Make it self-serve. Wrap models in simple endpoints, SDKs, or SQL functions so analysts and engineers can self select the magic without a helpdesk ticket. 

• Instrument & iterate. Track performance drift, cost, and business lift; retrain or retire ruthlessly based on evidence. 

• Teach the village. Run demos, share code snippets, and mentor teammates on pragmatic ML patterns that survive first contact with customers. 

 

Preferred experience (great to have, but not required) 

• End to end ML product ownership—from prototype notebook to cloud native service 

• Fluency in Python with libraries such as scikitlearn, PyTorch, TensorFlow, XGBoost, LightGBM 

• Experience choosing and finetuning foundation/LLM or diffusion models when they’re the quickest path to value 

• Comfort with feature stores, vector databases, and MLOps stacks (Airflow/Prefect, MLflow, Kubeflow, SageMaker, Vertex, or equivalents) 

• Both batch and low latency serving patterns (REST, gRPC, or streaming) 

• SQL that hunts for signal in messy data and A/B results 

• Solid grounding in statistics and experimental design, plus the storytelling chops to explain lift to non-data partners 

• Version control, CI/CD, and a bias toward shipping thin vertical slices over monoliths 

 

You’ll thrive here if you… 

• Think “impact > model elegance.” You pick the simplest approach that moves the KPI. 

• Prototype loudly. You’d rather show a working demo than a 40page deck. 

• Stay humble. If a spreadsheet baseline wins, you celebrate—and then raise the bar. 

• Translate effortlessly. You can chat GPU kernels at noon and revenue funnels at 12:05. 

• Love ambiguity. Blank whiteboards signal possibility, not paralysis.