Farisayi Dakwa

Built a product-matching system for cross-retailer price comparison, reaching over 90% accuracy by combining and fine-tuning sentence/vision-transformer embeddings with FAISS similarity search. Reduced retailer data acquisition from 20+ hours to under 3 hours per retailer by re-engineering Python browser automation with asynchronous parallel sessions.

Developed automated data pipelines in R, enhancing data processing efficiency and error detection. Created interactive visualizations and utilized statistical modeling to improve data accuracy and decision-making.

Built a vision-LLM document pipeline using Mistral, Qwen, and Gemini Batch API to extract structured tables from scanned PDFs. Created R Shiny and ggplot2 interactive dashboards and visualizations that became the company's standard reporting format.

Designed and implemented machine learning pipelines achieving high predictive accuracy. Processed geospatial data and created visualizations to communicate complex trends effectively.

Built predictive models of seabird density and occurrence from satellite-derived features and 2,284 ship-survey transects. Benchmarked four model families and combined them into a weighted ensemble, processing 50K+ geospatial data points and publishing results in CCAMLR Science.

Led analysis of a 25-year monitoring dataset on two penguin populations, quantifying declines and testing drivers. Applied time-series methods and published findings as first author in the African Journal of Marine Science.

Developed a machine learning pipeline for product classification and sales analysis, achieving high accuracy. Created an interactive dashboard to visualize sales trends.

Currently pursuing an MSc in Artificial Intelligence Engineering, covering machine learning, software design, CI/CD, and cloud architectures.

Completed the Machine Learning Specialization, covering machine learning fundamentals and model fine-tuning.

Completed the IBM Data Science Professional Certificate, covering data science methodologies and tools.

Completed an MSc by research in Biological Sciences, analyzing 25 years of population, diet, and ocean-climate data using trend models, regression, GAMs, time-series, and multivariate permutation methods.