image/svg+xml

Gaurav Koley code, research, writing, poems

I build production LLM infrastructure, evaluation systems, retrieval and agent workflows, and the operational machinery that makes AI products reliable, observable, and economically sane.

I also research language-model agents, social simulations, and programmable subjects.


Now

I am currently a Software Engineer, AI at Klarity. I am focused on systems that let teams ship AI features without turning every release into a model gamble: multi-provider execution, eval gates, replay-based migration, cost analytics, tracing, and incident workflows.

Research-wise, I am interested in LLM agents as experimental subjects: how they remember, coordinate, imitate, drift, and respond to simulated social environments.


Featured Open Source Projects

pdfvuer A PDF viewer for Vue using Mozilla's PDF.js that supports both Vue2 and Vue3 949 ★
saadhn A desktop client for JioSaavn, based on the official JioSaavn web app. Built with Electron. 66 ★
sani Less Is More. Less Code, More Blogging. 15 ★

View All

Recent Publications

LLM Agents as Programmable Subjects: Assays and Benchmarks for Agentic Behavior and Alignment

Gaurav Koley and Aditya Thiruvengadam

SALM: A Multi-Agent Framework for Language Model-Driven Social Network Simulation

Gaurav Koley

View All

Recent Talks

2020

2019

View All

Recent Posts

What an agent-native browser should own

If an agent can act inside a browser, who should remember what it is allowed to do?

Why computer use alone is not enough

Once models learned to read screenshots and move a pointer, browser agents started to feel almost obvious. A person can use a website by looking at it and cl...

View All