Alexey is currently unavailable

Alexey Zankevich

Verified Expert in Engineering

Machine Learning Developer

Harrington Park, NJ, United States

Toptal member since July 11, 2012

Expertise

Python Microservices Development Kubernetes Datadog Machine Learning Recommendation Systems PostgreSQL PyTorch OpenCV Apache Airflow gRPC Cloud Engineering Django

Bio

Alexey is a machine learning engineer with experience in model development, ML infrastructure, and real-time inference (Python microservices running in Kubernetes). He focuses on recommendation systems in Peloton. Alexey is the founder of Perfect Label, a machine learning startup.

Portfolio

Peloton Interactive

Python, PyTorch, Machine Learning, Recommendation Systems, gRPC, PySpark...

Tipsi

Python, Machine Learning, Django, OpenCV, Management

Experience

Python - 15 years
Microservices - 10 years
Machine Learning - 9 years
Python Asyncio - 7 years
PyTorch - 6 years
Kubernetes - 5 years
Apache Airflow - 5 years
Recommendation Systems - 5 years

Preferred Environment

Python, PyTorch, PySpark, Apache Airflow, Django, Python Asyncio, PostgreSQL, Kubernetes, Microservices, Machine Learning

The most amazing...

...accomplishment I've had is the Perfect Label Recognition Engine (Perfectlabel.io).

Work Experience

Senior Machine Learning Engineer

2018 - PRESENT

Peloton Interactive

Led the recommendation system for the newly released product (the rower).
Led the home screen recommendations (the main screen) for connected fitness devices.
Led the personalization microservices (the real-time serving part of our recommendations) based on Python, Asyncio, and Kubernetes.

Technologies: Python, PyTorch, Machine Learning, Recommendation Systems, gRPC, PySpark, Apache Airflow, Datadog, Microservices

CTO

2013 - 2018

Tipsi

Joined a very early-stage startup and turned it into a company with a very strong technological stack.
Made critical decisions regarding the architecture and managed a team of around 11 people.
Designed the recommendation system in the wine market.

Technologies: Python, Machine Learning, Django, OpenCV, Management

Experience

Perfect Label

I am the founder of this ML-based startup.

PRODUCTS
• Reverse wine search by label pictures
• Label unwrapper
• Label stitcher

These products are based on Airflow pipelines, OpenCV for image processing, and PyTorch models for feature extraction.

Unwrap Labels Library

https://github.com/Nepherhotep/unwrap_labels

A library to unwrap labels using the 6-point method. Based on six points (four corners and two middle points), it builds a mesh around the bottle and flattens the tiles to create a plane surface. Despite the library not having any neural networks embedded, it can easily be used with one if you have a custom model.