AI & Data Science.

The client needed a search solution for a large database of scientific medical articles used by physicians to gain insight on a specific topic, case, or condition. Traditional search methods tended to yield results with thousands of articles, making it nearly impossible to find useful information without many hours of research.

Using natural language processing and unsupervised machine learning techniques, we created an AI solution that clusters thousands of articles from search results based on their content. In addition, we built a sophisticated topic modeling mechanism to name the clusters and implemented several options to control the clustering process, making our AI tool flexible and highly intuitive.

Our clustering solution drastically improved the research process. By saving physicians significant amount of time and effort, it became one of the main tools of the platform, helping physicians find hidden gems in the vast ocean of medical science publications.

We developed a DVC-based framework for automatic training, validation, evaluation, and deployment of DL models for localization and classification of plant diseases with versioning of the main code, model parameters, and learning dataset, including a tool to automate the label collecting process of hard-to-label data from a group of labelers.

Current applications of this technology include plant disease management and growth phase detection and optimization.