AI Training Data Producer / RLHF & Red Teaming Lead (Founder & CEO, Precifarm AI Ltd)
As the founder of Precifarm AI Ltd, I led the evaluation of agentic AI systems with a practitioner-first focus on model outputs in engineering and IoT domains. I designed and implemented RLHF workflows, preference ranking, step-by-step chain-of-thought verification, and adversarial red teaming for alignment. My work included expert annotation, technical rubric creation, and production of high-quality training data for applied machine learning models. • Produced training data for technical and engineering problem/answer pairs, including counterfactual and edge-case generation. • Implemented comparative evaluation and DPO pair generation for advanced reward modeling. • Crafted adversarial prompts and conducted red teaming for model safety and bias testing. • Performed multilingual evaluation and rubric-based assessment in English, Swahili, and French.