Nihar Turlapati

About

Built production LLM/RAG systems during a UPS internship, including a shipment knowledge agent used across 15+ hubs worldwide and a multi-agent PDF RAG workflow. Stands out for combining hands-on enterprise integration with rigorous evaluation, hallucination reduction, and efficient fine-tuning techniques like LoRA.

Key Strengths

• Built production RAG agent at UPS for shipment knowledge retrieval across 15+ hubs worldwide
• Architected multi-agent workflows with ingestion, retrieval, and answer orchestration
• Strong grounding and hallucination mitigation in LLM systems
• Designed structured tool calling with JSON schema validation and retry handling
• Applied LoRA fine-tuning to reduce overhead by nearly 40%
• Built offline eval frameworks using recall, MRR, and faithfulness metrics
• Improved shipment update performance by 18%
• Reduced hallucination rate by 20% through hierarchical context packaging and citation constraints
• Experience integrating LLM agents into enterprise N-tier and streaming systems
• Handled messy operational data and privacy-sensitive shipment information

Experience

Education

Skills

Languages