ELVIS MAINA

Developed and optimized real-time video analytics pipelines for object detection and tracking using YOLO, TensorFlow, and OpenCV, improving overall detection accuracy from 51% to 69%. Created computer-vision training datasets, evaluated AI outputs for hallucinations/bias/safety, and designed prompts for language and vision-language tasks.

Developed and optimized real-time video analytics pipelines for object detection and tracking, improving overall model quality and detection accuracy from 51% to 69%. Created computer-vision training datasets and evaluated AI outputs to identify hallucinations, bias, and safety issues, providing feedback to improve model performance.

Built text analytics workflows to evaluate and refine AI responses for accuracy, tone, and relevance, including linguistic evaluation for Swahili. Engineered predictive forecasting models (XGBoost, LSTM, Prophet) and automated data pipelines with Python and Selenium, conducting adversarial testing for edge cases and system biases.

Used text analytics to process and evaluate large datasets, refining AI model responses for accuracy, tone, and relevance, including Swahili linguistic structural evaluation. Built predictive ML models for economic and financial forecasting (EURUSD, XAUUSD) and implemented automated Python/Selenium data pipelines with adversarial testing and guideline-based QA.

Bachelor of Science in Data Science at Cooperative University of Kenya (expected graduation 2027), with coursework in advanced statistics, machine learning, data engineering, text analytics, and computer vision.

Pursuing a Bachelor of Science in Data Science at Cooperative University of Kenya, with an expected graduation in 2027. Relevant coursework includes Advanced Statistics, Machine Learning, Data Engineering, Text Analytics, and Computer Vision.