At Lawcentral, I refactored a production AI engine into a configuration-driven architecture, enabling dynamic reasoning and reusable backend workflows. I also integrated configurable rule injection into a multi-agent AI system.
I worked with a parallelized Playwright evaluation harness for LLM-powered workflows, using deterministic validation and custom Server-Sent Events synchronization. I also debugged issues across backend services, asynchronous events, and model-driven workflows.
At Sabudh Foundation, I developed a multi-agent RAG system using Llama 3, LangChain, and Pinecone for question answering over uploaded documents. I optimized retrieval and inference workflows and built a Streamlit interface for the pipeline.
In my projects, I built multi-agent systems for code review, document retrieval, and AI interviews. My Multi-Agent RAG Framework reduced query latency from 3.5 seconds to 1.2 seconds while maintaining 95% retrieval relevance.
