Ram Kottala

About

Built a production internal LLM-powered knowledge assistant using a RAG architecture (Python, LLM APIs, cloud services) that answers employee questions with sourced, grounded responses from internal documents. Demonstrates strong practical depth in retrieval tuning (chunking/metadata filters), orchestration with LangChain, and production reliability practices (latency optimization, automated embedding refresh, evaluation metrics, logging/monitoring) while partnering closely with non-technical operations teams.

Experience

Education

Key Strengths

• Built and deployed an internal LLM/RAG knowledge assistant end-to-end (ingestion, chunking, embeddings, vector search, response generation)
• Improved retrieval quality through chunk-size/overlap experimentation and metadata-based filtering
• Implemented hallucination reduction by constraining responses to retrieved documents (grounded answering)
• Optimized production pipeline for lower latency and faster response times
• Set up automated ingestion and embedding refresh workflows to keep the knowledge base up to date
• Designed structured LLM workflows using LangChain to improve maintainability, debugging, and scalability
• Established evaluation approach with test prompts, golden datasets, and metrics (accuracy, relevance, latency, failure rate) plus production logging/monitoring
• Collaborated with non-technical operations stakeholders via requirements gathering, demos, and feedback loops