Aryan Kekre

I’m an AI/LLM Analyst focused on making LLM outputs dependable. I evaluated and scored 100+ LLM-generated outputs daily for accuracy, coherence, and task alignment across multiple model workflows.

I designed prompt evaluation frameworks to surface model failure modes, reducing response inconsistencies through refined annotation rubrics. I also annotated, validated, and quality-controlled training datasets that contributed to measurable improvements in model performance.

Earlier, I worked as a Data Annotation Associate, annotating 1,000+ football match video frames with bounding boxes and action labels to support computer vision training for sports analytics. I maintained annotation accuracy above project quality benchmarks under strict labeling guidelines.

Alongside this, I’ve gained enterprise workflow exposure as an SAP Consultant Intern, operating SAP Inventory Management modules under mentorship and documenting processes to identify system optimization opportunities. My projects reflect my engineering drive—building a production-style identity document verifier with a transformer architecture and a 3D-aware RAG platform using FastAPI, ChromaDB, and Gemini, with an interactive Streamlit evaluation workflow.