Anatolii Rubanov

- Built production AI pipelines for biomedical knowledge graph development and technical standards digitization. - Unified 100+ datasets, research papers, patents, APIs, and relational sources into a 600M+ RDF-triple knowledge graph for scientific research. -Improved extraction accuracy from 60–65% to 92% and scaled processing from 100 to 10,000+ engineering standards. - Built validated RDF outputs, LLM evaluation workflows, and delivered critical graph-to-visualization integration on deadline.

- Improved poster image-matching accuracy by 17% using CLIP, HDBSCAN clustering, and centroid-based search, reducing manual corrections. - Replaced fragile OpenCV template matching with a ResNet detection pipeline, improving UI-change robustness and reducing codebase size by 26%. - Maintained smart TV UI scrapers and screenshot-processing apps, sustaining a 95%+ scan success rate across 20+ platform-region combinations. - Built automated app name and logo recognition from live UI screenshots using VLMs.

- Built and trained a neural network classifier combining text embeddings, numerical metadata, and movie crew collaboration graph features, achieving 0.93 ROC AUC. - Collected and processed metadata for 6,000–7,000 movies, creating BERT-based emotion/sentiment features, description embeddings, and FAISS-based clustering.

Promoted from Research Intern in August 2019. - Designed custom experimental equipment, including a high-temperature XRD chamber, and conducted specimen preparation, XRD/mechanical testing, data analysis, and research reporting. Tech Stack: Python, XRD Data Analysis, Mechanical Testing Analysis, FEM Modeling, Scientific Computing

5-year specialist program, equivalent to a combined Bachelor’s and Master’s degree.