At Fractal Analytics, I built a DeepEval chatbot evaluation suite with threshold-based scoring for faithfulness, answer relevancy, hallucination, and context precision and recall. Across three release cycles, its RAG tests helped reduce hallucinated responses by approximately 30%.
I also traced LLM requests in Langfuse across model calls, tool usage, and retrieval steps on Azure OpenAI deployments, cutting debugging time from hours to minutes.
For application testing, I independently built a Python Playwright framework from scratch using the Page Object Model, covering approximately 600 test cases. I added reusable Pytest utilities and Pandas-based data validation, and used MCP and AI agent capabilities for bug recording and self-healing.
Earlier, at IMARC Services Private Limited, I automated data pipelines and integrated FastAPI with PostgreSQL for an internal application. At Tata Consultancy Services, I designed UiPath automations that saved 75 hours monthly across global business processes.
