Muhammed Rizwan

Designed a persona-driven RAG chatbot using GPT-4o and Azure AI Search to automate exploration and analysis of customer persona data for business stakeholders. Generated synthetic customer personas from unstructured customer success stories using Few-Shot and Chain-of-Thought prompting, benchmarking quality–efficiency trade-offs. Conducted statistical evaluation (McNemar’s test) across completeness, relevance, and consistency metrics to compare prompting strategies and guide system refinement. Augmented the chatbot knowledge base with synthetic personas, improving average response accuracy from 5.88 to 6.42 and achieving 81.82% stakeholder-reported business usefulness.

Built and deployed a RAG-based coaching chatbot using LangChain, Pinecone, and Streamlit, delivering an end-to-end GenAI application deployed on Streamlit Cloud. Collected and processed content from 30+ sources, implementing chunking and embedding pipelines to improve semantic retrieval quality. Evaluated multiple embedding models and LLMs through manual response comparison, refining prompts to improve answer relevance and consistency.

Installed and configured IBM FileNet P8 on Linux, ensuring stable document management operations across enterprise environments. Investigated application logs and performance metrics to diagnose root causes and resolve system-level issues, maintaining a 98% SLA resolution rate. Monitored application performance using Dynatrace and collaborated with cross-functional teams to improve system stability and release coordination.