Principal Data Scientist - Deep Learning
At Jampp, we’re on the mission of playing a leading role enabling the mobile app economy to grow. How? We build technology to support the most ambitious companies - from gaming to commerce pioneers - to propel the reach of their apps and a…
jampp · На сервисе с: 17.09.26 18:19
WHO WE ARE
At Jampp, we’re on the mission of playing a leading role enabling the mobile app economy to grow. How? We build technology to support the most ambitious companies - from gaming to commerce pioneers - to propel the reach of their apps and accelerate their mobile businesses.
We solve the most complex and large-scale technological challenges in mobile advertising today. We process over 2,500,000 ad requests per second, which amounts to over 300TB of data per day, across three global data centers. We rely on real-time machine learning models (with billions of features) that give us predictions in less than 100ms.
WHY DO WE NEED YOU?
Data, and how it is used, plays a central role at Jampp and is at the heart of how we make product, business, operational, and financial decisions.
Our Data Science team tackles challenging problems across many technical disciplines, including time series forecasting, graph mining, algorithmic optimization on petabytes of data, causal inference with missing data, and machine learning at scale.
We are now taking our machine learning platform to the next level. We are evolving from models built around manually designed features towards a deep learning architecture based on embeddings, allowing our models to learn richer representations of users, devices, creatives, publishers, advertisers, apps, campaigns and placements.
As a Principal Data Scientist – Deep Learning, you will play a key role in defining and leading this transformation. You will provide technical leadership across the development of our next-generation machine learning architecture, from the design of deep learning models and embedding strategies to their deployment and evolution in production.
You will work hands-on on some of the most challenging modeling problems in programmatic advertising, while setting technical direction, establishing best practices and helping other Data Scientists and engineers make sound architectural and modeling decisions.
This is not a research silo. You will have the opportunity to work with massive, high-cardinality datasets, real-time bidding systems and models that directly influence how Jampp evaluates and bids on millions of advertising opportunities every second.
WHAT YOU’LL DO
- Define and drive the technical strategy for Jampp’s deep learning and embedding-based modeling architecture.
- Design, develop and iterate advanced DNN architectures for prediction and optimization, initially focusing on CPI/CPA use cases and progressively expanding into real-time bidding, bid optimization, ranking and campaign optimization.
- Define the architecture and strategy for learning rich representations of high-cardinality entities such as users, devices, creatives, publishers, advertisers, apps, campaigns and placements.
- Lead the design of reusable embedding and representation-learning approaches that can support multiple models and use cases across the platform.
- Identify opportunities to improve the performance, scalability and generalization of our machine learning models using raw signals and learned representations.
- Develop modeling approaches for real-time prediction and decision-making in programmatic advertising, where models operate within strict latency and scale constraints.
- Evaluate new modeling approaches and technologies, balancing state-of-the-art deep learning techniques with the practical requirements of large-scale DSP and RTB systems.
- Establish technical standards and best practices for model development, experimentation, evaluation and productionization across the Data Science team.
- Guide complex modeling initiatives from problem definition and experimentation through production deployment and continuous improvement.
- Design, code and deploy machine learning models and supporting production tools, primarily in Python, remaining hands-on with the most technically challenging parts of the work.
- Work closely with ML Engineers, Data Engineers and Software Engineers to shape the training infrastructure, data pipelines, feature infrastructure, serving architecture and feedback loops required to support the next generation of Jampp’s ML platform.
- Analyze model and product performance metrics to understand how algorithmic changes impact bidding decisions, campaign performance, user response and business outcomes.
- Provide technical mentorship and guidance to Data Scientists, helping them navigate complex modeling problems and develop stronger approaches to experimentation and model design.
- Communicate technical findings, architectural decisions, trade-offs and recommendations clearly to both technical and non-technical stakeholders.
- Collaborate with Data Science and ML teams across Jampp and Affle to identify opportunities for shared capabilities, knowledge and machine learning solutions.
REQUIREMENTS
- Significant experience in Data Science, Machine Learning, Deep Learning or a closely related quantitative/technical role, with a track record of leading complex machine learning initiatives.
- Strong academic background in Computer Science, Applied Mathematics, Physics, Statistics, Engineering, Econometrics, or another quantitative field.
- Deep understanding of machine learning and deep learning fundamentals, including neural network architectures, representation learning, optimization and model evaluation.
- Extensive hands-on experience developing and deploying Deep Learning models in production environments.
- Strong experience with Python and the scientific/machine learning Python ecosystem.
- Experience working with large-scale datasets and high-cardinality categorical or ID-based features.
- Proven track record of taking machine learning models from experimentation and research through reliable production deployment.
- Experience designing or making significant technical contributions to ML architectures, training pipelines, model serving or other machine learning infrastructure.
- Experience working with real-time or latency-sensitive machine learning systems, ideally in advertising, marketplaces, recommendations or other high-throughput environments.
- Strong analytical and problem-solving skills, with the ability to independently investigate ambiguous problems and define effective technical solutions.
- Strong technical communication skills, with the ability to influence modeling and architectural decisions across multidisciplinary teams.
- Experience providing technical leadership, mentorship or direction to other Data Scientists or engineers.
- Comfortable conducting daily professional communications in English (written and verbal).
YOU MAY BE A GREAT FIT IF…
- You have designed and deployed deep learning systems based on embeddings, representation learning or other approaches for learning from high-cardinality entities.
- You have experience working on DSPs, programmatic advertising, real-time bidding (RTB), ad exchanges, ad networks or other real-time advertising systems.
- You have experience applying machine learning to bidding, bid optimization, CTR/CVR prediction, ranking, campaign optimization or other decision-making problems in advertising.
- You have experience with recommendation systems, ad-tech, pricing, ranking, personalization, fraud detection, or other large-scale optimization and prediction problems.
- You have worked with systems processing very large volumes of ad impressions, auction events or real-time user and publisher signals.
- You have experience modeling highly cardinal entities and complex interactions across users, devices, apps, creatives, publishers, advertisers, campaigns or similar entities.
- You have experience designing or evolving ML platforms, training pipelines, feature stores, model serving or monitoring systems.
- You have worked with models operating at very high scale and under strict latency constraints.
- You have experience balancing model complexity and predictive performance with computational cost, scalability and production constraints.
- You have a strong track record of turning research ideas into production systems and measuring their real-world impact.
- You enjoy defining technical approaches to problems where there is no obvious answer and where modeling decisions can have a significant impact on the product and the business.
- You are comfortable providing technical leadership while remaining hands-on with the most challenging modeling and engineering problems.
- You are able to influence technical decisions through strong reasoning, experimentation and communication rather than relying solely on formal authority.
- You like working in a self-sufficient, autonomous manner, striving through ambiguity and taking ownership of complex technical problems.
- You have a strong sense of urgency and ownership over the product, and care deeply about the quality, scalability and impact of the solutions you build.
- You are curious, pragmatic and comfortable balancing technical depth with practical business impact.
- Smarts, humility, and equal willingness to learn and teach.
Похожие вакансии Data Science & ML
Управление Модельных Рисков - «by-design» уникальное подразделение, которое «видит» все AI/ML модели Банка, понимает процессы/точки принятия решений моделями, кривая обучения у наших сотрудников «круче», чем в подразделениях, разрабатывающих модели.
Наша цель – помочь Банку зарабатывать больше и улучшать клиентский опыт за счет повышения эффективности моделей и их использования в Бизнес процессах.
Мы ищем DS/Quant с экспертизой на стыке финансовой математики, торговых алгоритмов и ML для работы в напралении Модельного Риска Торговой Книги. Вы будете выступать ключевым звеном контроля качества моделей и агентов, используемых в оценке рисков и ценообразовании финансовых инструментов.
Обязанности
- оценка качества моделей ClassicML, работающих на финансовых рынках, в том числе: торговые алгоритмы, оценка эластичности по цене, кластеризация клиентов и т.д.
- проведение независимой оценки качества моделей прайсинга финансовых инструментов, а также риск-моделей (включая VaR, CVA, PFE)
- анализ корректности математических предпосылок, устойчивости моделей и их соответствия бизнес-требованиям и рыночным условиям
- оценка и контроль рисков агентов, работающих в торговых алгоритмах, процессах прайсинга финансовых инструментов и оценки рыночных рисков
- создание инструментов для автоматизации процессов мониторинга моделей и агентов с использованием Python
- подготовка аналитических заключений по результатам валидации, формирование рекомендаций по доработке моделей
- презентация результатов работы, аргументация выводов и защита предложенных решений перед разработчиками моделей, риск-менеджментом и бизнес-заказчиками.
Будет плюсом:
- навыки работы с генеративными AI-моделями; опыт создания AI-агентов и использования их в работе будет преимуществом.
Требования
- знание теории вероятностей, математической статистики и случайных процессов
- опыт работы на стороне разработки или валидации ClassicML моделей
- уверенные навыки программирования на Python, опыт разработки структурированного и поддерживаемого кода
- знание базовых принципов построения моделей прайсинга производных финансовых инструментов
- опыт работы с количественными моделями в финансах или смежных областях будет преимуществом
- участие в летних школах или курсах: Vega, ЦМФ – также является преимуществом.
Условия
- работа в удобном и современном офисе недалеко от станции метро Кутузовская
- ежегодный пересмотр зарплаты, годовая премия
- корпоративный спортзал и зоны отдыха
- более 400 образовательных программ СберУниверситета для профессионального и карьерного развития
- расширенный ДМС, льготное страхование для семьи и корпоративная пенсионная программа
- гибкий дисконт по ипотечному кредиту, равный 1/3 ключевой ставки ЦБ
- подписка Прайм с возможностью совместного использования на трёх близких
- вознаграждение за рекомендацию друзей в команду Сбера.
HireSeeker собирает вакансии со всех площадок и присылает только релевантные. Бесплатно.