data scientist for personalization

HireHi; United States; Remote; USD 130,000 - 180,000; Full time

Job summary

HireHi ищет специалиста в области машинного обучения и анализа данных для построения и развёртывания продакшн‑моделей, работающих на пропенсити, ценообразование и персонализацию. Ваша задача — привести модели от идеи к эксплуатации через этапы формирования задачи, обучения, оценки и мониторинга.

Кандидат будет сотрудничать с командами продукт и инжиниринга, работать над экспериментальными платформами, обеспечивая соответствие API контрактам и низколатентный сервис на GCP.

Qualifications

• Опыт разработки и развёртывания ML-моделей в продакшн.
• Умение работать с Python и инструментами ML (scikit-learn, XGBoost/LightGBM).
• Сильные навыки SQL и работы с dbt/BigQuery.
• Опыт causal inference, uplift modeling или экспериментальной аналитики.
• Опыт в 4–7 лет в роли data scientist/ML engineer.
• Свободное владение английским языком.

Responsibilities

• Разрабатывать, валидировать и разворачивать продакшн-модели ML для предсказаний.
• Сопровождать жизненный цикл моделей: от формулировки задач до мониторинга.
• Работать с A/B тестами для измерения эффекта на метриках.
• Разрабатывать low-latency serving на GCP и API контракты.
• Строить пайплайны предсказаний и мониторинг качества.

Skills

Python

scikit-learn

XGBoost/LightGBM

SQL

GCP/AWS/Azure

dbt

CI/CD

Causal Inference

English

Production ML

Education

CS/Math degree

Tools

Git

BigQuery

dbt

GCP

Job description

Описание

Timeleft is a social networking platform.

Задачи

Build, validate, and ship production ML models for propensity, pricing and discount optimization, personalization, churn, and LTV Own the model lifecycle from problem framing and feature engineering through training, evaluation, deployment, monitoring, and retraining Write tested, versioned, reviewed production code alongside Engineering Design and ship a personalized discounting model that selects offers for users at the paywall Partner with Product on A/B tests and holdouts to measure causal lift Build a measurement framework for pricing and discount decisions Establish low-latency model serving on GCP Define which features should be precomputed in BigQuery/dbt and which require fresh, real-time pipelines Set up monitoring for drift, staleness, and prediction quality Translate product problems into modeling problems and model outputs into API contracts for Engineering Document handoffs so Engineering can maintain the serving layer independently Design uplift and causal models to identify users who respond to discounts Run and interpret experiments measuring models’ incremental impact on revenue and retention Design an experimentation program for improving data science and machine learning models Deliver a personalized discounting model in production for live paywall decisions Establish a documented, reusable low-latency serving pattern on GCP Demonstrate incremental lift on a core business metric through a controlled experiment Put model monitoring and drift and staleness alerts in place Create a repeatable model-to-production playbook for the team

Требования

Strong Python skills for data science and ML, including scikit-learn and XGBoost/LightGBM Proven experience shipping models to production, including at least one model serving live traffic Hands‑on experience with a cloud ML platform, ideally GCP, or ability to transfer equivalent AWS/Azure experience Solid SQL and experience working with a dbt/BigQuery warehouse Software engineering fundamentals: git, code review, testing, and CI/CD Experience with causal inference, uplift modeling, or applied experimentation 4–7 Years in a data scientist or ML engineer role, including personally taking at least one model from prototype to live production serving real users or real traffic Quantitative background in CS, statistics, engineering, or equivalent hands‑on experience Fluent English Strong focus on business metrics, ability to communicate with Engineering and Business, explain modeling tradeoffs in business terms, and work with ambiguity

Будет плюсом

PyTorch or TensorFlow, streaming/event pipelines such as Pub/Sub, Dataflow, or Kafka, pricing, discounting or personalization, feature stores or versioned feature pipelines, multi‑armed bandits or reinforcement learning, startup experience, B2C, subscription, or marketplace experience

Условия

Async case study assessment as part of the recruitment process