OJHero M.
Mid-Level AI Trainer
Open to offersNew to PlatformAbout
Mandisa M. is an experienced AI Trainer with a Bachelor of Science from the University of Cape Town, specializing in the training, evaluation, and enhancement of large language and multimodal AI models. At Sama Ai. and Soul AI., Mandisa engaged in cross-domain AI instruction across STEM, biology, humanities, sports, and multilingual environments, creating high-quality training examples and conducting multifaceted assessments focused on reasoning assessment, safety policy evaluation, and dataset quality assurance. Mandisa also provided structured feedback to address hallucinations, reasoning gaps, and instruction-following failures, ensuring model robustness and alignment. In the audio domain at Babel Audio., Mandisa performed tasks such as audio annotation, transcription, and speaker segmentation for speech recognition and language understanding models, consistently maintaining annotation accuracy in multilingual datasets.
Experience
AI TRAINER
Sama Ai.Trained and evaluated large language models across various domains, including STEM and multilingual subjects. Created complex training examples and conducted thorough assessments on factors such as factual accuracy and ethical alignment.AI MATHEMATICS & BIOLOGY TRAINER
Soul AI.Specialized in training AI models for scientific reasoning, evaluating problem-solving approaches, and ensuring the clarity of scientific explanations. Developed domain-specific evaluation criteria and strategies for addressing troubleshooting.AUDIO ANNOTATOR
Babel Audio.Annotated and performed quality checks on audio datasets for models focused on speech recognition and language understanding. Handled tasks such as transcription, segmentation, and speaker identification while maintaining annotation consistency across multilingual samples.
Skills & Expertise
Education
- Bachelor of ScienceUniversity of Cape Town