Sriram Ramesh

I’m an AI Data Annotator and LLM Quality Analyst focused on large language model evaluation and multimodal dataset validation. I benchmark model outputs for truthfulness, helpfulness, factual accuracy, safety, and instruction adherence, using structured, rationale-driven QA.

In my current Data Annotator role, I perform image and video annotation with detailed visual descriptions, frame transformation analysis, and entity tagging. I create and refine bounding boxes to improve computer vision dataset accuracy, and I review transcription accuracy to strengthen quality assurance.

I also evaluate LLM outputs across empathy, verbosity, contextual relevance, and PII compliance, and I simulate AI chat interactions involving text, image, and video generation tasks. By comparing multiple model outputs and documenting structured evaluation rationales, I help drive clearer training feedback and support reinforcement learning workflows.

As a bilingual expert (English–Tamil), I simulate real-time conversational scenarios using two Gemini model variants and benchmark results with evaluation rubrics. I translate and localize content, assess cultural appropriateness and semantic similarity, and recommend updates that improve the Tamil locale experience.