Piyush Rajendra

About

Built and deployed a GPT-4 + Pinecone RAG system that lets users query large internal document collections with grounded, cited answers. Demonstrates strong applied LLM engineering (chunking experiments, hallucination controls, metadata recency boosting) plus production-minded evaluation/monitoring and performance tuning (rate-limit mitigation via pooling/batching). Also effective at translating complex AI concepts to non-technical stakeholders through prototypes and live demos, helping secure client sponsorship.

Experience

Education

Key Strengths

• Built and deployed a production RAG system for querying large document collections
• Practical mitigation of LLM context window limits via tested chunking strategy (256/512/1024; selected 512)
• Hallucination reduction through citation grounding with stored source metadata and constrained generation
• Operational ownership: weekly monitoring of latency/accuracy and systematic review of failed queries
• Workflow orchestration using LangChain chains integrating retrieval, GPT-4 generation, and citation formatting
• Performance optimization under vector DB rate limits using connection pooling and batch queries
• Reliability approach using clear success criteria and real-world A/B testing
• Stakeholder communication: explains LLM/RAG concepts (hallucinations, vector DBs) with demos and simple examples to secure client buy-in