Faaiz Kamal

Agentic AI Teacher at Amazon

Create and annotate multimodal training data (text, image, video) to improve large model performance across technical domains. Automate data workflows using Python and SQL to increase annotation efficiency and reduce manual processing. Produce structured outputs across diverse reasoning styles to improve model instruction-following and consistency.

Generalist AI Trainer at Mercor

Recorded end-to-end task demonstrations across multiple applications while providing real-time reasoning commentary to capture decision-making processes. Completed complex workflows (e.g., video editing, multi-tool operations) to generate high-quality process supervision data. Performed comparative evaluation of model outputs, rating responses across dimensions such as accuracy, instruction-following, clarity, and usefulness.

AI Fellow at Handshake AI

Annotated video data using masking and bounding box techniques for object removal and reconstruction tasks. Designed prompts to guide video-based model transformations, including removal and reintegration of subjects. Evaluated model outputs for accuracy, completeness, and visual coherence.

Generalist AI Trainer at Abaka AI

Evaluated AI-generated image edits using structured scoring frameworks across multiple quality dimensions. Developed and applied rubrics for instruction-following and visual consistency, improving evaluation reliability. Provided detailed justifications for ratings to support model training and alignment improvements.

Generalist AI Trainer at Micro1

Evaluated robotic interaction outputs for realism and physical consistency in simulated environments, adapting quickly to evolving guidelines and dynamic evaluation standards. Identified and categorized model errors, creating structured annotations and bounding boxes to localize failure cases. Served as a quality reviewer, auditing annotations and acting as a final validation layer before client delivery.

AI Tutor at xAI

Delivered high-accuracy (95%+) data annotation and evaluation for large-scale AI model training. Improved annotation workflows through prompt and instruction refinements, reducing turnaround time by 20%. Built a custom Chrome extension to track performance metrics and optimize team productivity.

AI Trainer at Outlier

Designed, reviewed, and fine-tuned prompts across generalist tasks, STEM subjects, preference ranking, audio-to-text transcription, and image-to-text interpretation. Solved self-created and existing prompts, delivering detailed step-by-step feedback on AI outputs to enhance model reasoning, problem-solving capabilities, and understanding of complex queries.