Mar 2020 – Feb 2021 · 11 mos · New York, United States

• (Vpipe is a cloud-based serverless AI model serving. It is an MVP-stage project developed by two people.)
• Created performance tool for automatic model inference and instance configuration
• Developed Grpc-based Python SDK with shared memory module for maximum performance

Nov 2018 – Jan 2020 · 1 yr 2 mos · Remote

• Developed Machine Learning solutions in large-scale traffic analysis system
• Designed and implemented cascade pipeline for 1000fps per GPU Car Detection and Tracking
• Trained and optimized Object Detection and Recognition models with Tensorflow
• Conducted research in Person Re-Identification in Complex Environments (topic of master’s thesis)

Jul 2018 – Sep 2018 · 2 mos · Greater New York City Area

• Implemented request grouping for better traffic analysis at Google Bigtable service
• Created request groups analysis tool from scratch

Jul 2017 – Sep 2017 · 2 mos · Zürich Area, Switzerland

• Introduced Fibers for better concurrency at latency-critical service(called for ~50% of all search requests)
• Produced production performance measurements, shared experience with the team

Jul 2016 – Jun 2017 · 11 mos · Moscow, Moscow City, Russia

• Automated 5-10% of text dialogues for SME, Individuals and other departments
• Led chat suggestions feature with the team of 5 persons; Reported results to the VPs
• Conducted research in Reinforcement Learning for Dialogue Systems (topic of bachelor's thesis)
• (Tinkoff is the largest Russian online-only bank. It has >500 chat operators and >4000 at call-center)

Jul 2015 – Mar 2016 · 8 mos · 530 University Avenue, Palo Alto

• Back-end(Ruby on Rails) + Front-end(JS, Jquery, Kefir.js, etc)