At FireFlink, I built a Python framework that turns natural-language test steps into executable Selenium scripts, supporting workflows across Web, Android, and iOS. I also developed locator-generation and fallback workflows to recover from failed element identification.
I reduced the framework’s end-to-end processing time from about eight seconds to about two seconds per workflow. That’s a 75% latency reduction.
I built a RAG pipeline that generates structured test scenarios from application documentation. Using LlamaIndex, LangGraph, Qdrant, ChromaDB, and Redis, I developed document retrieval and stateful workflows for test-case generation.
I also contributed to FireFlink’s human-in-the-loop automation platform, developing Kafka- and WebSocket-based workflows and integrating multimodal inputs and multiple LLM providers. As an AI Intern at FireFlink, I built an AI-powered Resume Builder and Code Translator.
