Shahbaz Shaik
building the plumbing for LLM agents
I work on LLM infrastructure and agent systems — the layer that sits between a model and something actually useful.
What I’m working on
Most of my time goes into Jeeves Core — a runtime for LLM agents.
It started as a simple question: once you get a local model running, what does it actually take to make it useful? The answer turned out to be everything around the model — persistent memory, logging, config, database access, clean module boundaries. The agent frameworks I tried kept tangling these concerns together with the agent logic itself, and rewriting that plumbing for every new idea got old fast.
So Jeeves Core is the layer underneath. Agents plug into a shared runtime that handles the boring but load-bearing parts, and each capability only has to care about its own job. The Code Analysis Pipeline and Personal Assistant are two capabilities built on top of it — different problems, same foundation.
The broader thing I care about: the bottleneck for useful agents isn’t model quality anymore, it’s everything else. That’s where I want to spend my time.
Featured Projects
Jeeves Agentic Runtime
A runtime that lets LLM agents pause, resume, and share infrastructure
Jeeves Code Analysis Pipeline
An agent that explores codebases without making things up
Jeeves Personal Assistant
A capability built on the Jeeves runtime — tasks, notes, and long-term memory