---
date: '2026-05-23'
description: consulting on machine learning systems, inference engines, and model behaviour
id: consult
layout: letter
modified: 2026-06-06 23:54:26 GMT-04:00
tags:
  - evergreen
title: consulting.
created: '2026-05-23'
published: '2026-05-23'
pageLayout: letter
slug: consult
permalink: https://aarnphm.xyz/consult.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
My consulting [[thoughts/craft|work]] is in [[thoughts/Machine learning|machine learning]] systems and efficient inference engines. I also study [[research|model behaviour]]. This [[index|website]] is the non-physical me, my notes, and my interests. I mainly work on open source, and I plan to keep doing so. As of 2026-09-06, it has 881,573 [[thoughts/|words]], counted by [[thoughts/scripts/calculate_tokens.py|this script]].

I focus on kernel optimization for LLM inference and on deployments where latency and throughput are measured end to end. I offer the following services:

- custom prefill and decode kernels for latency-sensitive or high-throughput LLM deployments
- distributed LLM serving, including NCCL setup and debugging

If you would like to work with me, contact me at <services@aarnphm.xyz>.

![[thoughts/craft#open source|open-source work]]