pdf

Roman Glushko

Machine Learning Engineer

Profile

  • Aspiring problem solver & ML engineer
  • 10+ years of experience in software engineering
  • 3+ years of experience in technical management of scrum teams
  • Expertise in a broad variety of technical and business topics
  • Constant, quick & curious learner
  • Positive attitude

Skills

  • Experience: Tabular Data Processing, Image Processing, Classification, Regression, Time Series Forecasting, Object Detection, Data Analysis
  • Languages: Python, Golang, WebDev Stack (JS/ReactJS/TypeScript & HTML5 & CSS3)
  • ML Frameworks: PyTorch, XGBoost, Keras, Scikit-Learn, HuggingFace, langchain
  • MLOps: Kubernetes
  • DataViz & Processing: Pandas, Seaborn, Plotly

Education

  • Machine Learning Nanodegree, self-paced, 2020-2021
  • M.S. Computer Engineering, Khmelnytskyi National University, Ukraine, 2018

Projects

  • [Kaggle] Shopee - Price Match Guarantee

    The Shopee marketplace has a price match guarantee program. They needed a model that could automate the process of finding marketplace lots with similar product offerings and prices, despite variations in marketing strategy, product titles, and images used by merchants.

    Ranked in the top 45% of participants on Kaggle. Got gold and silver medals for providing deep insights on the domain and problem goal.
    • Experience: eCommerce, Tabular & Image Data, Python, Keras, Data Processing, embeddings
    • Links: • Competition

  • [Kaggle] HPA - Single Cell Classification

    Human Protein Atlas organization was interested in finding patterns of protein distribution inside of the single cell. This helped to understand the role of different proteins in disease development, medical treatment impact, etc.

    Based on 150Gb of 4-layers human cells images, it was needed to segment each cell and predict in which organelle of the cell the protein of interest was located. Provided dataset contained only inaccurate image-level labels (not directly connected to particular cells).

    Got to Top 15% of submitters on Kaggle.
    • Experience: Biology, Image Medical Data, Weakly-Supervised Learning, Data Analysis, Data Processing, Instance Segmentation, Multi-Label Classification
    • Links: • Competition

  • Rock, Paper, Scissors Game

    Trained a model to recognize hand forms of rock, paper, scissors and incorporate it into a game. Collected hundreds of images in order to improve the model accuracy.

    Ended up with a model that is 90% accurate and recognizes common ways to shape RPS forms. Deployed the model as a serverless web app using TensorFlow.js.
    • Experience: Problem Framing, Data Collection, Image Data, Python, Tensorflow, Tensorflow.js
    • Links: • Live Demo • GitHub
Other projects can be found on GitHub and Kaggle

Work Experience

  • Senior Software Engineer (R&D, AI Apps, Notebooks), DataRobot; 2021-Present (2+ years)
    • Designed and implemented an internal framework for building ML-enabled applications using the DataRobot platform and modern Python stack (Flask, Pydantic, FastAPI).
    • Built a complaint review app for a US healthcare company, contributing to the complaint analysis pipeline and NLP modeling using Hugging Face Transformers.
    • Deployed and troubleshot ML models in Kubernetes-based environments
    • Prototyped a content-aware conversational agent using OpenAI, LangChain, and Qdrant
    • The rest of my software engineering experience could be found there
    • Experience: Problem Solving, Python, Software Architecture, System Design, Distributed Systems, NLP, Huggingface Transformers, Machine Learning, Kubernetes
  • eCommerce Software Developer, Tech Lead, Atwix; 2015-2020 (5.5 years)
    • Experience: eCommerce, Problem Solving, Management, System Design, Problem Framing, PHP, MySQL, Elasticsearch, Redis

Management & Communication

  • Held a full-time technical leader position for 3.5 years (5-15 people teams incl. SEs, QA, PM, DevOps) + seasonal leading of several internal projects and initiatives
  • Write technical blog posts and thoughts about management. Organized technical & business newsletter (led for 2 years).

Hobby & Interests