This job has closed.

AdZeta · 4 days ago

Founding ML Engineer (Ads & Signal Engineering) - NYC/ SF Preferred

New York, NY

Full-time

Hybrid

Mid, Senior Level

AdZeta is building the signal engineering and activation layer for modern growth teams, focusing on unifying first-party data and predicting customer LTV. They are seeking a Founding ML Engineer to design and implement user-level pLTV models and a signal pipeline for adtech applications.

Computer Software

Responsibilities

Design + ship LTV prediction systems end-to-end: data → features → training → evaluation → deployment → monitoring

Build modeling approaches that work under adtech realities: delayed conversions, sparse signals, noisy attribution, changing platform policies

Own MLOps: reproducible training, CI/CD for models, logging/monitoring, data quality checks, drift + performance alerting

Partner with product + GTM to translate business outcomes (profit, payback, repeat rate) into model objectives and activation strategies

Help define our 'Signal Engineering' playbook: what we compute, how often, and how we send it

Qualification

LTV modelingProduction MLPythonSQLMLOpsAdtech experienceCloud environmentsDockerBigQueryScikit-learnAirflow

Required

Strong ML foundations + applied instincts (you can reason from objective → data limitations → model choice → deployment)

Hands-on experience building production ML (training pipelines, deployment, monitoring; not just notebooks)

Experience with LTV modeling (e.g., probabilistic/BTYD-style thinking, survival/retention modeling, regression/classification for value, calibration)

Excellent Python + SQL; comfortable in modern data stacks and cloud environments

Preferred

Adtech/Martech experience: Meta CAPI, Google Ads/Enhanced Conversions/Offline Conversions, audience/CRM activation, conversion quality, incrementality intuition

Identity/data joining experience (hashed PII, multi-key matching, deduping, event stitching)

Experience with streaming/near-real-time systems and event pipelines

Familiarity with experimentation frameworks (uplift/incrementality), MMM/attribution constraints, or measurement-heavy environments