mirror of
https://github.com/hanzoai/papers.git
synced 2026-10-29 07:34:31 +00:00
zeekay
f9db4f5737
Bylines, paper authorship, correspondence addresses and the founder blurb under each post named an individual. They name the organisation that published the work instead: the brand on posts and pages, the research byline on papers, and research@ in place of a personal address. Both are what these repositories already used elsewhere — nothing new was coined. Co-authors are untouched. Where a paper named someone else alongside, they stay named; only the one byline changed. Bibliography entries, running heads, PDF metadata and affiliations move with it, so a citation still resolves to whoever the cited paper names. Sample and fixture data uses the placeholder name these tests already use elsewhere, so no assertion changed meaning.
68 lines
3.1 KiB
TeX
68 lines
3.1 KiB
TeX
% Hanzo Platform: PaaS for AI-Native Application Deployment
|
|
\documentclass[11pt,twocolumn]{article}
|
|
\usepackage{../shared/hanzocover}
|
|
\usepackage{amsmath, amssymb, amsthm}
|
|
\usepackage{mathtools}
|
|
\usepackage{bm}
|
|
\usepackage{graphicx}
|
|
\usepackage{booktabs}
|
|
\usepackage{hyperref}
|
|
\usepackage{enumitem}
|
|
\usepackage{algorithm}
|
|
\usepackage{algpseudocode}
|
|
\usepackage{xcolor}
|
|
\usepackage{float}
|
|
\usepackage{subcaption}
|
|
\usepackage{multirow}
|
|
\usepackage{listings}
|
|
\input{../shared/lstlang}
|
|
\usepackage[numbers]{natbib}
|
|
\hypersetup{colorlinks=true,linkcolor=black,citecolor=blue,urlcolor=blue}
|
|
|
|
\newtheorem{theorem}{Theorem}
|
|
\newtheorem{lemma}[theorem]{Lemma}
|
|
\newtheorem{definition}{Definition}
|
|
\newtheorem{proposition}[theorem]{Proposition}
|
|
|
|
\lstset{
|
|
basicstyle=\ttfamily\small,
|
|
breaklines=true,
|
|
frame=single,
|
|
numbers=left,
|
|
numberstyle=\tiny,
|
|
xleftmargin=2em,
|
|
}
|
|
|
|
\title{Hanzo Platform: PaaS for AI-Native Application Deployment}
|
|
\author{
|
|
Marcus Chen \quad
|
|
David Wei \quad
|
|
Hanzo AI Research \\
|
|
\textit{Hanzo AI Research} \\
|
|
\texttt{[email protected]}
|
|
}
|
|
\date{December 2021}
|
|
|
|
\begin{document}
|
|
|
|
\hanzocoverpage
|
|
|
|
\begin{abstract}
|
|
We present \textbf{Hanzo Platform}, a Platform-as-a-Service (PaaS) designed specifically for deploying AI-native applications---systems that incorporate LLM inference, vector databases, model serving, and GPU compute as first-class primitives. Traditional PaaS offerings (Heroku, Vercel, Railway) treat AI workloads as an afterthought, requiring manual configuration of GPU instances, model serving infrastructure, and inference APIs. Hanzo Platform introduces three innovations: (i) an \emph{AI-aware build system} that automatically detects model dependencies, optimizes container images for inference workloads, and configures GPU scheduling, reducing deployment time for AI applications from hours to minutes; (ii) a \emph{unified resource model} that treats LLM API access, vector database provisioning, GPU allocation, and traditional compute/storage as a single resource plane with declarative configuration; and (iii) a \emph{cost-optimized autoscaler} that dynamically scales inference endpoints based on token throughput, request latency, and cost budgets, achieving 38\% lower inference cost compared to static provisioning. We evaluate Hanzo Platform against five competing PaaS offerings on deployment time, operational cost, scaling responsiveness, and developer experience across 15 representative AI application architectures. Results demonstrate 4.7x faster deployment, 38\% lower cost, and 2.1x better scaling responsiveness. Production deployment data from 14 months of operation with 2,800+ applications and 47M inference requests validates the approach.
|
|
\end{abstract}
|
|
|
|
\input{sections/01-introduction}
|
|
\input{sections/02-architecture}
|
|
\input{sections/03-ai-aware-build-system}
|
|
\input{sections/04-unified-resource-model}
|
|
\input{sections/05-cost-optimized-autoscaler}
|
|
\input{sections/06-developer-experience}
|
|
\input{sections/07-security-and-compliance}
|
|
\input{sections/08-evaluation}
|
|
\input{sections/09-production-deployment}
|
|
\input{sections/10-related-work}
|
|
\input{sections/11-discussion}
|
|
\input{sections/99-conclusion}
|
|
|
|
\end{document}
|