Roman Glushko

I’m a Machine Learning and Deep Learning Engineer with nearly a decade of experience in Software Development. For me, engineering is a pool of extremely challenging tasks that I enjoy solving. That’s why I started my way from e-Commerce where I had been working for more than five years solving challenges for multiple US- and EU-based B2C and B2B businesses. Three years out of five, I was leading several full-stack cross-functional agile teams and closely worked with my clients on identifying their challenges and solving them through software development.

Then, I have moved to an even more challenging area where I’m currently building an innovative AI platform solving open-ended R&D problems of an enterprise product with my ex-Google, Microsoft, Meta, Cisco, Twitch colleagues.

Finally, science, Machine Learning and Deep Learning are taking a special place in my life. These are the most challenging fields and the most capable tools that humanity came up with. So I get a special excitement when I see a reasonable use case to apply them in my day-to-day job.

In the meanwhile, I’m putting my thoughts and experience together and sharing them through technical articles on my website.

Roman’s Experience

As an ML/DL engineer, I’m proficient in engineering software in modern Python (Flask, FastAPI, Pydantic, asyncio). My AI stack includes MLFlow, LabelStudio, Airflow, XGBoost, Sklearn, PyTorch, Huggingface Transformers, Pandas, Seaborn. All that goes hand in hand with applied statistics and ability to read and use ideas from novel research papers. I can do data analysis, plan and execute A/B tests.

Also, I have a great experience of designing and developing asynchronous distributed microservice multilanguage systems and deploying them in a reliable, resilient and secure way using AWS, Kubernetes, Helm, gitOps operators. I usually use Elastic Stack with Logstash, Prometheus and Jaeger as an observability platform.

As a secondary stack, I use Golang, NodeJs, Socket.IO and ReactJS with Typescript to tackle problems where Python is not the best/right choice.