Middle Data Scientist (Prom.ua) in EVO

Middle Data Scientist (Prom.ua) Kyiv, remotely Prom.ua is Ukraine's largest marketplace, selling over 100 million products from tens of thousands of entrepreneurs across the country. On Prom.ua: every buyer can find everything they need at the best price: from a toothbrush to a cultivator for a garden. every entrepreneur can sell goods in the catalog of the marketplace, on the website created on the Prom platform and in the "Prom shopping" mobile application. Prom.ua in numbers: 4.8 million people visit the marketplace every day, and more than 60,000 people work on the marketplace. companies in the catalog of 120 million products About the Data Science team: We optimize various parts of the product using data and machine learning algorithms. In parallel, we are building AI systems that provide a strategic business advantage and move the company in the direction of e-commerce of the future. Now there are 5 people in the team: 4 Data Scientists and Team Lead. Areas of work of the team: Product recommendations and personalization Search and ML-ranking Machine translation of product content Automatic moderation of products in the catalog, classification of products Identification of product duplicates Generation and validation of tags for SEO Peculiarities of working in a team: High involvement in the product environment, close inter-team interaction — > little research goes under the table , many models in production Understanding of the set goals, orientation on the result -> models do what is necessary, do what is not necessary Absence of bureaucracy, the opportunity to participate in the selection of tasks, a developed culture of initiative and responsibility for the result Focus on infrastructure development for greater reliability of decisions, automation of routines and creation of new opportunities for tasks Collaboration and teamwork spirit: mutual concern and support, friendly atmosphere Exchange of experience: author's courses, project presentations, team grooming, etc. We build close ties with the development team. Analysts help us make a business assessment of decisions. For day-to-day work, a JupyterHub server is installed with the ability to set the necessary characteristics of the working environment, you can work on a local machine if necessary. We have our servers with video cards for teaching and deploying models. Technological stack: Programming language: Python Data analysis and processing: Jupyter Notebook, Pandas, NumPy Machine Learning and Deep Learning: Scikit-learn, TensorFlow, PyTorch, FAISS, XGBoost Data visualization and monitoring: Matplotlib, Seaborn, Plotly, Bokeh, Tableau, Grafana Databases : Postgres Big Data and distributed computing: Apache Spark, Hadoop MLOps: MLflow, DVC, TensorFlow Serving, Python packaging, Fast API Dags: Airflow Data queues: Kafka Search: Elasticsearch. For this role, it is important: a deep understanding of neural networks, especially in NLP, experience working with frameworks for developing neural networks (pytorch/tf), experience working with machine learning: problem formulation, data collection and research, model training, evaluation of results, analysis of model performance, preparation for deployment; experience in deploying and maintaining a model in production, improving existing models, the ability to write reliable and clean code in Python, understanding and using various data structures, OOP, etc.boatingVC (Git etc); experience in working with databases, SQL queries, willingness to dive deeply into business problems and translate them into ML terms (architecture, loss functions, metrics) To be a plus: experience in writing neural networks from scratch, describing results and research, experience in learning models of data that exceed the amount of memory, experience with highly loaded systems, Big Data and distributed computing, experience in applying MLOps practices: control of code versions, data and models, automatic deployment, monitoring of logging, model testing, retraining of models, experience of working with zembeding Tasks: improvement of the system for finding duplicate products, improvement of the product classification model and machine translation, development of new recommendation models systems, generation of more conversion content for products, research of new areas of application of machine learning to solve business problems Selection stages: Meeting with a recruiter and tech leader Technical interview with project engineers Final interview with Head of Data Science Prom.ua We offer: Official employment of the company's staff. 24 calendar days of paid vacation per year, unlimited sick leave. Remote work. The possibility to visit the office of the Kyiv. Medical Insurance. Services of a corporate psychologist.