Danylo P.

Team Leadership & Strategy: Manage a team of 5 Data Scientists and 1 Data Engineer; oversee hiring, performance reviews, and strategic roadmapping aligned with company-wide field-zone objectives. Fiscal Responsibility: Directed $20k budget for GPU infrastructure and $12k for cloud service optimization to support high- compute model training. Agri-Tech Impact: BBCH Classification: Led end-to-end development of Transformer-based models for crop growth stages, enabling precise pesticide application and preventing chemical waste. Process Automation: Standardized corn/sunflower kernel counting using MobileNet/MobileViT; reduced counting time by 8x while increasing data accuracy over manual field methods. Developed model for offline, on-device inference with a highly optimized 5MB footprint, ensuring high performance in remote areas with no connectivity. Strategic Decision Support: Developed yield prediction models for wheat, corn, and sunflower, serving as critical data points for Traders and the Trade Committee. Facilitated the delivery of a large-scale agrologistic optimization model, managing the project lifecycle and cross- departmental coordination to achieve an estimated $1M in annual operational saving Operations: Spearheaded a data-driven culture by collaborating with DevOps to implement Azure-based CI/CD pipelines with daily/weekly automated model retraining loops.

Product Vision: Led the development of an MVP for bark beetle detection using satellite imagery and Transformer architectures. Infrastructure: Engineered a robust Data Pipeline and LoRaWAN infrastructure within AWS to support real-time IoT and satellite data processing. St rategic Research: Partnered with national research institutes to curate and label high-resolution satellite datasets for forest health monitoring.

End-to-End Vision Pipeline: Took full responsibility for the design, training, and deployment of a high-speed UI/UX recognition service. Model Orchestration: Developed an infrastructure that inference mobile screen and processed them through a high- capacity ResNet-based architecture to deliver precise bounding-box labels for UI elements.