Fahad Mattoo

I'm a Senior Software Engineer at Emumba in Islamabad, where I work on AI agent systems and large language model-powered applications.

At Emumba, I'm building Avtar, an avatar-based personal AI assistant, and Chatterbox, a multi-agent chatbot system. My work spans LangGraph-based agentic workflows, multi-agent orchestration with ~20 integrated tools, real-time voice processing with LiveKit, computer vision pipelines using MTCNN and rembg, and memory retrieval optimization with Mem0 and pgvector. I'm also implementing RAG pipelines for legal document processing — including contract analysis and document hierarchy detection — as well as integrating TTS services and the Hedra API for avatar face generation.

Previously, I worked as a Machine Learning Engineer at DCube Tech, where I built RAG-based chatbots, a multi-user multi-document chat engine, and an end-to-end structural prediction system for API call payloads. Before that, I worked as a Data Scientist at PackageX on Named Entity Recognition pipelines, as a Research Assistant at NCRA-RDDL applying traditional ML techniques across diverse domains, as a Teaching Assistant at NUST-EME in the Mechatronics department, and as a Management Associate at CBL overseeing the Mechanical Workshop department.

Interests

I'm deeply passionate about the evolving landscape of Generative AI and autonomous systems. My current areas of interest include:

• AI Agents & Multi-Agent Systems — designing goal-driven agents that reason, plan, and collaborate to solve complex tasks using frameworks like LangGraph and LangChain.
• Model Context Protocol (MCP) — building and integrating MCP servers to extend agent capabilities with external tools and data sources.
• Agentic RAG — moving beyond static retrieval pipelines toward agents that adaptively query, reason over, and synthesize information from multiple knowledge sources.
• LLM Orchestration — exploring patterns for memory management, structured outputs, tool use, and long-horizon planning in production-grade LLM systems.

Deep Learning Setup