Parth Sareen

I work on Ollama and lead the agent tooling around it.

On the side, I build small tools for myself to experiment and push agents to their limits. Before that, I built a startup on agent reliability and worked at Apple, Tesla, and more.

Outside of that: latte art, Muay Thai, and writing down ideas as they form.

writings

Jul 14 · 16 min Deep dive
Building a harness for open models

What open models expose about context, tools, compaction, and the agent loop around them.

Read the deep dive →
Jun 15 · 5 min Essay
The philosophy of Ollama Launch

A design diary for Ollama Launch: simplicity, invisibility, and the small moments that make keeping up with AI easy.

Read the essay →
Mar 25 · 8 min Essay
Yet another generalist

Quick brain dump. Getting 70–80% good at anything, why specialists might still win, and the strange grief of not-quite-writing code anymore.

Read the essay →
Dec 25 · 15 min Deep dive
Building reliable AI agents

A first-principles breakdown of agent loops, tool design, and context engineering.

Read the deep dive →
Sep 10 · 10 min Deep dive
Sampling and structured outputs in LLMs

How sampling shapes token selection, and generating structured outputs in modern models.

Read the deep dive →

recommended readings

Loading readings…

elsewhere

Small tools

watchy · ducky · zuko

Hobby

Latte art

Hobby

Fragrance collection

Public talks and interviews are collected here.