Senior applied scientist (data science & ml)

Company Description

Ad Zeta turns insights from your data lake to profit signals that seamlessly integrate with your ad stack. We unify your first-party data, predict lifetime value, activate high-value signals, and prove revenue impact— governed, auditable and real-time. Ad Zeta was built on the premise that data without activation means nothing.

Role Description :

We’re not hiring a model‑tuner. We’re hiring a builder‑owner . The ideal candidate loves taking ownership, accountability and thrives in 0 → 1 execution.

You'll own the entire  end‑to‑end ML pipeline and orchestration layer — from raw data → features → models → deployment → monitoring → activation. You’ll ship scrappy v1s fast, then harden to reliable v2s. Zero‑to‑one, then one‑to‑many. Have prior experience working in adtech / martech (highly preferred).

What you’ll do

Data plumbing & schemas :  Ingest and model data across Shopify / GA4 / ESP / CRM / ad platforms; define event taxonomy, identity graph, and data contracts. Feature engineering :  Build offline / online features for p LTV, propensity, and clustering; ensure parity with a feature store (e.g., Feast) and model registry (e.g., MLflow). ML pipeline & training :  Stand up reproducible training pipelines (Python / SQL / dbt) with experiment tracking, hyperparameter search, and evaluation. MLOps & deployment :  Package models (Docker), deploy to managed services (Vertex / Sage Maker / Cloud Run), wire CI / CD (Git Hub Actions) and blue‑green / rollback patterns. Orchestration :  Schedule jobs with Airflow / Prefect; manage dependencies, SLAs, retries; design backfills and incremental loads. Observability :  Implement DQ checks (Great Expectations / dbt tests), model performance & drift monitoring, alerting, and cost guardrails. Activation & APIs :  Ship fast, robust services (Fast API / Cloud Functions) to push signals into Meta CAPI, Google Ads (EC / OCI), Tik Tok Events API  or via reverse ETL (Hightouch / Census). Experimentation & measurement :  Partner with product / GTM on uplift tests, holdouts, and calibration to prove incremental value. Docs & enablement :  Produce crisp runbooks, ERDs, and playbooks; teach others to self‑serve.

30

• 60
• 90 day outcomes

30 days :  Audit current data + tracking; publish V1 tracking plan & data model; baseline p LTV / propensity notebook and EDA; quick wins on data reliability. 60 days :  Production warehouse models (dbt) + orchestrated pipelines; first model in staging with offline / online validation; reverse ETL wired for a design partner. 90 days :  v1 serving in prod with monitoring; value‑based bidding pilot live in Google / Meta  for 2+ partners; dashboarding for KPIs and model health; draft case study.

You’ll be great at this if you

Have 7-10+ years  shipping DS / ML systems to production (not just notebooks). Are fluent in Python, SQL , and one major cloud (GCP / AWS ), with Big Query / Snowflake  and dbt  experience. Know MLOps  (MLflow / Weights & Biases, Feast / feature store, Docker / K8s, CI / CD) and orchestration (Airflow / Prefect). Can design measurement plans and communicate trade‑offs with clarity.

Thrive in ambiguity, move fast, and own outcomes end‑to‑end.

Nice to Have :

Streaming (Pub / Sub / Kafka), privacy / consent (Consent Mode v2, GDPR / CCPA), clean rooms (ADH / Meta), Shopify / subscription analytics.

Why Ad Zeta?

Founding‑team impact, meaningful equity , and the chance to build the ML backbone of a category‑defining product.

How to Apply?

DM with : / Git Hub, 2–3 bullets on your best 0→1  ML build (problem → approach → impact), and a link to a repo / case study (if public).