Aditya Arya

Improved production RAG retrieval relevance by 20–25% by redesigning the ranking stage with query classification and hybrid scoring (Reciprocal Rank Fusion + normalized vector/BM25 combination). Prevented silent quality regressions in production by building an automated evaluation suite on RAGAS Faithfulness and Context Precision metrics, run before every deployment. Shipped a bilingual (English/Hindi) fine-tuned assistant using Gemma-3-1B-IT via Unsloth; resolved multi-turn context loss with custom memory tracking and state management.

AI Developer Intern - AmberFlux EdgeAI Private Limited

(2025-06 - 2025-08)

• Improved production RAG retrieval relevance by 20–25% by redesigning the ranking stage with query classification and hybrid scoring (Reciprocal Rank Fusion + normalized vector/BM25 combination).
• Prevented silent quality regressions in production by building an automated evaluation suite on RAGAS Faithfulness and Context Precision metrics, run before every deployment.
• Shipped a bilingual (English/Hindi) fine-tuned assistant using Gemma-3-1B-IT via Unsloth; resolved multi-turn context loss with custom memory tracking and state management.

Bachelor of Technology - Computer Science and Engineering - Rajiv Gandhi Institute of Petroleum Technology (2023 - 2027)