Senior Data Analyst (AdTech / AI)

UPPeople

Рівень:: senior
Джерело:: djinni.co

Відгукнутись на вакансію →

Що робити

Measurement & Evaluation Foundations: Define how we measure what matters — data quality, the accuracy of our AI scoring, how many records actually get matched to usable signals, cost per outcome. Build the feedback loops that tell us whether a change makes the system better or worse.
Scoring Intelligence (co-owned with engineers): Partner with the engineering team on the LLM-based scoring system. You will build the evaluation framework for AI scoring, run experiments comparing prompts, models, and approaches against a baseline, and make data-backed calls on where we invest next. Our AI scoring system is actively evolving — you will shape how we measure and improve it.
Client-related Analysis: Own the analytical side of the work that internal sales and media-buying teams bring in from pharma clients — how reachable a target audience is, how good the data looks, why a campaign performed the way it did. You turn their questions into data-backed answers they can use. Identity is a product core; the analyst works with internal teams, not with clients directly.
Cost and Value Analysis: Turn cost and value into product decisions. Which data sources and enrichment scenarios pay off, where we overspend, what to cache, what to cut. We operate a mix of free owned data and paid external providers — the decisions have real dollar impact.
Active AI Usage: Use Claude Code and modern AI engineering tools to multiply your own output. Automate the repetitive parts of analysis so you can focus on the decisions. We provide Claude Code licenses and expect active daily use.

Що очікуємо

5+ years as a Data Analyst, Analytics Engineer, or Data Scientist in a product team (not consultancy, not BI-only roles).
Advanced SQL at TB scale — BigQuery, Snowflake, or equivalent warehouse. You can read and write complex queries on tens of terabytes without someone pre-modelling the data for you.
Python for analysis at a working level — pandas, Jupyter, or modern equivalents. Comfortable putting together a small script or helper when an existing tool doesn't cover your case.
Built measurement or evaluation frameworks from scratch for systems that had no baseline. You are comfortable defining what and how to measure, not consuming someone else's KPI.
Comfort with probabilistic systems. You understand that noise is inherent and 100% accuracy is not the goal — scoring, matching, ranking.