Ahmed Dhankwala

Experience

1. Llm Trainer Annotator (Evaluation & Multimodal Annotation)

   Turing

   Nov 2024 — Present

   Created high-quality datasets for LLM supervised fine-tuning (SFT) and RLHF, generating prompts, evaluating outputs, identifying failures, and producing gold-standard corrections. Designed diverse prompt sets and test cases to analyze model reasoning, safety alignment, robustness, and edge-case behavior. Performed QA by auditing and validating other annotators’ work, ensuring accuracy, consistency, correct model-break classification, and adherence to client SOPs. Identified recurring model weaknesses and provided structured correction signals to support reward modeling and iterative model improvements. Produced multimodal video annotations, segmenting actions with timestamps and detailed reasoning to support training of vision-language models (VLMs). Consistently met strict quality benchmarks and deadlines across multiple confidential AI development projects.