Raman05 Thakur

I’m a backend developer and AI agent engineer focused on building production-ready AI agent systems, LLM orchestration layers, and autonomous workflows. With 1.5+ years of experience, I enjoy turning complex capabilities into reliable, tool-using agents that can operate across long-running sessions.

In my current role as Lead Developer / Lead architect for ZenithOS, I’m architecting a modular AI agent framework where specialized organs coordinate autonomously to execute system-level tasks. I built the Cortex Kernel for central reasoning and mission-critical resource allocation, and I manage stateful context using vector databases for memory across long-running sessions. I also integrate toolsets via Model Context Protocol (MCP) for autonomous CLI and filesystem orchestration.

Previously at Labellerr, I built production LLM applications and AI agent systems for a leading data annotation platform. I developed a core Python SDK used by 500+ developers, created a RAG-based documentation engine with 92% retrieval accuracy (reducing support overhead by 35%), and fine-tuned/evaluated 200+ LLM/VLMs through benchmarking pipelines. I bring a strong engineering mindset—rigorous evaluation, modular design, and rapid iteration—backed by 150+ technical articles on LLM architectures and production deployment.