About

Building real systems, not just demos.

I'm Arunkumar — an AI systems engineer focused on building agent-based applications that work reliably in production. I work with startups to design and implement LLM-powered systems, from multi-agent orchestration to RAG pipelines and real-time monitoring infrastructure.

My work sits at the intersection of software engineering and applied AI. I care about the parts most people skip: retry logic, context window management, structured output parsing, tool interface design, and observability. The problems that determine whether an agent system actually works or just demos well.

Before focusing on agent systems, I spent years building backend services and distributed systems. That background shapes how I approach AI engineering — with an emphasis on reliability, debuggability, and operational clarity over hype.

I work primarily in Go and Python for distributed systems and AI infrastructure. I’ve built production systems on AWS using Kubernetes, Postgres, and event-driven architectures. My focus is performance, scalability, and operational simplicity.

What I work on

  • Agent orchestration — designing execution loops, tool interfaces, and decision routing for multi-step LLM workflows
  • RAG pipelines — retrieval architecture, chunking strategies, and embedding pipelines for production knowledge systems
  • Observability — structured logging, token economics dashboards, and trace-based debugging for agent systems
  • Reliability engineering — retry strategies, circuit breakers, context window management, and graceful degradation

How I work

I prefer small, focused engagements where I can go deep on a problem rather than spread thin across many. I write about what I build — not to market, but to think clearly and share what I've learned with others solving similar problems.

If you're building something with LLMs and need someone who cares about the engineering as much as the AI, get in touch.