Egor Yatsishin

Engineering data pipelines to make laboratory experiments useful and available for analytics. Building workflows and development processes for a new DE team from the ground up. – Developing Python scripts for ETL pipelines – Doing full-cycle development of a dedicated app for report generation and laboratory operation speed enhancement – Leading the DevOps part of the DE team – Interacting with the laboratory to understand their needs – Mentoring colleagues and integrating common development practices

Creating a personal single-screen conversational assistant for the largest department store in Eastern Europe specialized on luxury goods. As the first developer, built and led linguistic development department. Conducted technical interview, onboarding, code-review and task distribution for backend developers.

Programmed backend on Kotlin for iOS app from scratch to pre-production state. Personally worked on creating client flow through bot states, developed widgets for in-app presentation of goods from TSUM stock, established API communication between services, managed codebase architecture.

Carried out an R&D project on assessing the need of using ML methods for coreference resolution on Russian medical forum data for the company needs. - Analyzed coreference strategies and linguistic aspects of Russian forum speech, identified suitable SOTA ML models, assessed existing relevant corpora. - Composed the pipeline from raw data to prediction scoring, processed and analyzed 1.7 million coreferential chains, adapted English BERT-based system to Russian language.

COLING is a major annual international conference on computational linguistics. I was leading and representing a team of 3 more students. - Giving a 10-minute speech as the winner of the workshop dedicated to extraction of adverse side-effects caused by pharmaceutical products from Twitter in Russian. - Achieved results were reflected in a publication.

Highly selective (10%) summer school at renowned Skoltech which covers the modern statistical machine learning methods. - Presenting a 15 minute talk on "BERT Implementation for Detecting Adverse Drug Effects Mentions in Russian".