Senior Machine Learning Engineer

• Jan 2017 - Present

Medialogia is a leading company in the Russian market performing collection and analytics of data from the media and social networks with a large number of commercial clients from all industries. In the course of work, I developed 5 ML and DL services that provide important functionality for the company's clients. Services work with Kafka, Rabbit and Redis, placed in Docker containers and run through Kubernetes or Docker-compose. The total throughput of services is more than 70 million messages per day. I provided a full development cycle - from the idea and initial hypotheses, including training models and writing production code. 1. Object sentiment service. It provides information for clients about customer's sentiment towards them and their competitors. (NLP, Tensorflow-Keras). 2. Spam detection service for social networks and media messages, which allowed to clear clients search results in our system from irrelevant advertisement and spam. (ML) 3. OCR (Optical character recognition) service, implementation of which provided the opportunity to enrich search results in our system for clients. (CV, Pytorch +TensorRT) 4. Object quotation detection service, which allowed us to provide information about quotations of their representative in the media for our clients. (NLP, Tensorflow-Keras). 5. The compilation of social network accounts into a virtual persons project, which included particularly an avatar face ID service(CV, Pytorch). This project provided the opportunity to offer our clients generalized information about their clientele. Big data analytics including developing data quality tests and anomaly search (Hadoop, Scala, Spark, Pyspark, ElasticSearch) Mentored ML engineers in both machine learning and bid data processing. Show less