Machine Learning Developer in Reading, United Kingdom

Most of my research was devoted to developing adaptive ensembles of machine learning models in real time. An important property of these ensembles is that at any time in the future, their performance will be close to the performance of the best model in this ensemble. These ensembles are highly adaptive to the newly arrived data, which is particularly important in the real-time setting. My research was published in journals such as Machine Learning, Data Mining and Knowledge Discovery, and Neurocomputing.

Application to monitor tests coverage in pandas that is used by its development team. The application is developed using Streamlit and is hosted on Heroku. The data is saved and downloaded from AWS S3.

The increasing connectivity of data and cyber-physical systems has resulted in a growing number of cyber-attacks. Real-time detection of such attacks, through the identification of anomalous activity, is required so that mitigation and contingent actions can be effectively and rapidly deployed.

I created a new approach for aggregating unsupervised anomaly detection algorithms, which is to be used by digital identity provider companies. I developed the prototype in Python. The preprint is available at arxiv.org/pdf/2010.03857.pdf

The project aims to deploy an ML classification model on Heroku using FastAPI. Data Version Control (DVC) on AWS S3 is used for data versioning. API tests and unit tests to monitor the model performance on various data slices were implemented and incorporated into a CI/CD framework using GitHub Actions.

The project aims to build a reproducible ML pipeline for estimating a property rental price using MLflow and Weights & Biases. New data comes every week, and the model needs regular retraining. An end-to-end reusable pipeline will enable an easy retraining process and reduce the time-to-production.

The goal of the project is to set up processes and scripts to retrain, redeploy, monitor, and report on the ML model that estimates the attrition risk of a company. The project implements automatic data ingestion, training, scoring, deploying, and diagnostics.

I developed a new approach for dynamic cyber risk estimation. This new method assesses the maximum number of hacking attempts with the desired confidence. I designed the prototype in R and published the research in Springer, which can be read at doi.org/10.1007/s10618-021-00814-z.