Hozefa B

I’m an AI & ML engineer who builds deployable AI systems that help robots move reliably in real-world environments. I focus on agentic AI and RAG infrastructure, while staying grounded in classical ML/DL and computer vision needs.

Over 2+ years at CYNLR, I built robotics-focused ML applications including tracking, multi-arm collision avoidance, and object tracking with depth estimation. I work across the GenAI stack (LangChain, LangGraph, MCP, and agentic AI) and also integrate robotics tooling (ROS and relevant industrial communication workflows). I’ve supported Fanuc, Doosan, and Flexiv deployments by building ML systems around them in an agile environment.

I’ve also led and shipped projects end-to-end—most notably an AI agent for multi-arm coordination and self-collision avoidance. By integrating a solution with the CyRo three-arm robotic system, I reduced application time by almost 50%. I built a quotation generator agent using an MCP server/client model and created a RAG infrastructure platform with distributed ingestion, semantic retrieval, hybrid search, citation grounding, and hallucination detection.

My work extends into practical data and perception pipelines, including curating multimodal conversation text for translation models and delivering stereo-vision solutions like the AUDI recess strip insertion problem with tight 1mm–2mm tolerances. I’ve implemented computer-vision inspection with OpenCV (bringing inspection time to 100 ms per part) and built robotics services with LabVIEW, C++, and industrial protocols.