Machine Learning Lead
2019 - PRESENTEnvoy Media Group- Created an in-house framework that does auto-ML for data and tasks specific to Envoy Media Group. We used the AWS stack and could train and deploy a new model within an hour without writing code.
- Contributed to feature our framework as a case study for Partner Success on AWS (https://aws.amazon.com/partners/success/envoy-media-toptal/).
- Trained, tested, deployed, and monitored machine learning models with Amazon SageMaker.
- Collaborated with the Envoy Media Group team on their long-term AI/machine learning strategy.
- Created a framework that supports easy model creation and lifecycle management, including monitoring and visualization. We have from 10s to 100s of live machine-learning models deployed in production.
Technologies: Amazon Web Services (AWS), Python, Amazon SageMaker, Artificial Intelligence (AI), Artificial Neural Networks (ANN), Gradient Boosted Trees, Data Science, Machine Learning, Docker, Jupyter Notebook, SciPy, XGBoost, Explainable Artificial Intelligence (XAI), Machine Learning Operations (MLOps), Natural Language Processing (NLP)Machine Learning Engineer and Consultant
2022 - 2022Things Inc.- Trained our own diffusion models using various approaches. These models were done on Google Colab notebooks using weaker GPUs and smaller datasets, but we managed to get things working and were ready for large-scale experiments.
- Investigated a large number of papers and codebases related to Denoising Diffusion Probabilistic Models (DDPM) and Denoising Diffusion Implicit Models (DDIM).
- Did research on competitors, such as Midjourney, to understand where the industry stands at the moment.
Technologies: Python 3, PyTorch, Jupyter, Jupyter Notebook, Google Colaboratory (Colab), Computer Vision, Amazon Web Services (AWS), Amazon SageMaker, Generative Adversarial Networks (GANs), Diffusion Models, DDPM, DDIM, Image ProcessingHead of Machine Learning
2020 - 2021Aisle3- Developed the product matching engine prototype, which matched the same product from different sellers using their images and description. Used OpenCV and ConvNet-generated image features, as well as vector index storage and search.
- Oversaw the long-term ML strategy for the company, deciding which directions are the most promising going into the future. Worked closely with other teams on the overall system architecture on AWS.
- Supervised a small remote team tasked with bringing the product matching engine into production. By the time I left the company, we had a product matching API and an internally-developed labeling tool using the API.
Technologies: Python, Amazon Web Services (AWS), PyTorch, Deep Neural Networks, Artificial Neural Networks (ANN), Artificial Intelligence (AI), Computer Vision, OpenCV, TensorFlow, Deep Learning, FAISS, AWS Fargate, Amazon S3 (AWS S3), FastAPI, SQL, Image Processing, Jupyter NotebookAmazon SageMaker Consultant
2019 - 2020Visably LLC (via TopTal)- Provided consulting to the client to migrate their on-premise ML solution into Amazon SageMaker.
Technologies: Amazon Web Services (AWS), Amazon SageMaker, Jupyter NotebookMachine Learning Engineer (Remote)
2019 - 2019Pirate Labs- Created a recommender system delivering marketing emails for a company with multiple, diverse clients. Each client is a shop selling different products, and the shops are very different from one another. We created a single system that works for all.
- Designed, implemented, and tested the recommender system. It was highly configurable and flexible, enabling it to effectively adapt to unique settings of each shop.
- Oversaw the testing and helped with launching the system into production.
Technologies: Recommendation Systems, Factorization Machines, Python, Machine Learning, Data Science, NumPy, Pandas, SQL, Jupyter NotebookCo-founder, CEO
2018 - 2019NVision LLC- Created a cloud-based infrastructure for crawling, indexing, and supporting an image database of tens of millions of images.
- Retrieved images from a database of tens of millions of images. Query images could be very heavily altered versions of the original.
- Enabled digital watermarking of images (prototype).
Technologies: PyTorch, TensorFlow, Keras, OpenCV, Python, C++, SQL, Image Processing, Jupyter Notebook, Generative Adversarial Networks (GANs)Developer of Recommender System (Freelance)
2017 - 2018Triskk.com- Helped prototype a recommender system.
- Created, tested, and tuned a prototype with Python.
- Implemented the system within AWS infrastructure and made it production-ready.
Technologies: Amazon Web Services (AWS), Git, Jupyter, NumPy, Python, SQL, Jupyter NotebookProduct Manager
2016 - 2017Armsoft- Served as the product manager for an in-house developed ETL.
- Participated in product creation from the start: design, implementation, testing.
- Oversaw client deployments and service monitoring.
Technologies: TFS, R, Visual Studio, .NET, C#Head of Research and Education Center
2012 - 2017Armsoft- Created a highly demanded educational program with more than 10 applicants for one position.
- Co-developed the overall strategy for the education center, including creating the curriculum, designing the admission process, and recruiting the lecturers. Managed 1-2 assistants who took care of day-to-day operations.
- Managed entrance exams (up to three rounds) with more than 300 applicants and more than 20 lecturers/TAs/colleagues being involved in different rounds.
- Researched and published in IEEE TKDE, currently ranked #1 by Google Scholar in the category "Databases and Information Systems."
Technologies: R