About

I’m Deepan Seeralan, a software engineer based in Greater Seattle Area. I work on backend systems specializing in storage. Over the past couple of years my focus has shifted more towards AI infrastructure — building RAG pipelines, working with vector databases, and figuring out how to get LLM-powered applications from prototype to production without the wheels falling off.

This blog started as a general space for things I was learning. It’s now pivoting to a focused series on applied AI and AI infrastructure — the practical, code-heavy kind. Less theory, more “here’s what actually happened when I tried this.”

What I’m Writing About

So far that’s meant posts on RAG pipeline design, vector database tradeoffs (Pinecone, serverless), and taking a document Q&A service from prototype to production with FastAPI. I’m also tracking how emerging standards like Google’s Open Knowledge Format are shaping how agents read and reason over content.

Where This Is Headed

The throughline going forward is applied AI and AI infrastructure: LLM serving and deployment, evaluation and observability (LLMOps), and agentic systems — tool use, multi-step reasoning, and the infrastructure to run them reliably. Expect more code-first deep dives and honest notes on what didn’t work, as I build more of this out.

Previously

Before the AI focus, I wrote about backend systems, distributed systems tools (ZooKeeper, protobuf), cloud deployments on GCP and Azure, and occasional book notes. Those posts are still here if you’re looking for them.

Get in Touch

The comment section on each post (powered by GitHub Issues via Utterances) is the best place for post-specific questions. For everything else, LinkedIn or Twitter/X work well.

You can also subscribe via RSS to get new posts as they come out.