Daisy Kosgei

Developed FMNR Index scores and modelled adoption impact on household resilience and food security using two-stage treatment models and HTE analysis. Delivered causal insights to support resilience policy targeting.

Led the development of an interactive dashboard visualizing the socio-ecological impact of Ipomoea hildebrandtii on rangelands. Mapped species spread, carbon loss, and pastoralism threats; contributed to national strategy insights.

Conducted statistical analysis (ANOVA, regression) on biochar field experiment data to evaluate its impact on soil moisture and seedling growth. Produced data visualizations and recommendations for land restoration.

Developed a production-level time-series model capable of predicting daily billable hours with 95% accuracy, optimizing workforce planning and ensuring efficient resource allocation. Migrated the entire organization’s data for over 1100 users from Google Sheets to a new data platform, ensuring a seamless transition and reproduction of all reports.

Implemented an automated calculation system for incentives for over 2,000 individuals, enabling seamless report recreation with 100% accuracy. Developed a weekly seating allocation plan for over 1,500 employees, optimizing space utilization and enhancing operational efficiency.

Performed spatial analysis of forest data using ArcGIS and ENVI, providing critical insights to support sustainable forestry projects. Analyzed evapotranspiration data to identify patterns and trends, supporting water resource management in arid regions.

Developed a Logistic Regression Model to assess stroke risk factors, creating actionable healthcare insights from sampled data provided by The Moi Teaching and Referral Hospital. This project involved statistical analysis, modelling, and computing using R to identify key risk factors, ultimately contributing to improved decision-making in stroke prevention and patient care strategies.

Executed end-to-end development of a production-level Job Recommending system using NLP, Cosine Similarity, and Tf-Idf to match 1000+ candidates to roles with high accuracy. This project involved data preprocessing, feature extraction, and model development to efficiently recommend the best-fit candidates for each role, reducing time for the recruitment process and improving hiring outcomes.

Applied graph technology (TigerGraphs) and machine learning to detect money laundering, flagging high-risk entities for further investigation. Collaborated on a project that identified UN-sanctioned individuals using proxy analysis, enhancing due diligence processes.