Developed a cloud-native multi-agent LLM system using Google A2A and MCP protocols, enabling chain-of-thought analysis and RAG pipelines for 5G core configuration. Increased retrieval accuracy by 26% by pivoting from vector databases to graph databases (Neo4j) and LangChain memory stores. Performed Supervised Fine Tuning, KTO, and GRPO algorithms with QLoRA PEFT Built MCP servers for LLM tool calling, automating CRD deployment, scaling cloud resources, accessed via HTTP and SSE protocols via FastAPI. Exposed context retrieval of structured and unstructured data using multi-modal, semantic, and lexical searches, integrating tools such as Docling. Developed containerized and backend services on Kubernetes and Crossplane environments with Docker. Delivered proof of concepts to executives, expanding technologies to company wide production teams. Managed two interns, helping them deliver through agile user stories, improving team velocity and code quality.

Developed Java-based control software for robot systems, focusing on modular architecture, sensor integration, and path planning. Collaborated with teammates using Git, code reviews, and iterative testing cycles to deliver competition-ready robot software.

Promoted after 3 months as a result of exceptional performance and general recommendations from superiors. Delegated tasks, managed the safe and cash balances, assigned break schedules, and handled conflict resolution while working additionally as a cashier.