¸Ô±¾ÊÓƵ

Research Computing and Data Services

The ¸Ô±¾ÊÓƵ Supercomputer supports faculty, students, and staff with high-performance and high-throughput computing (HPC/HTC), GPU acceleration for AI, private large language models, and more than 1.5 petabytes of on-campus storage. These resources support research, teaching, and exploration across disciplines.

Get Started

2,500
Intel & AMD CPUs
45
NVIDIA GPUs
16
NVIDIA H100 80 GB GPUs
1.5 PB
Distributed Storage

What is the ¸Ô±¾ÊÓƵ Supercomputer?

The ¸Ô±¾ÊÓƵ Supercomputer is a group of networked compute nodes, GPUs, and local and network storage that handle work far too large for single desktop computer: simulations that run for weeks, AI models that need hundreds of gigabytes of GPU memory, and datasets measured in terabytes. The ¸Ô±¾ÊÓƵ Supercomputer brings undergraduate-friendly research computing to a liberal arts college.

Research

Computational science, genomics, ecology, chemistry, mathematics, data science, digital humanities and more.

Teaching

Course environments for statistics, data science and AI ethics, consistent for every student in a class.

Innovation

Custom web applications, APIs and research tools hosted and maintained on campus.

Available Computing Resources

The ¸Ô±¾ÊÓƵ Supercomputer pairs thousands of CPUs and dozens of NVIDIA GPUs with 1.5 PB of distributed storage. Everything is reachable through the campus network or the ¸Ô±¾ÊÓƵ VPN.

Compute Modes

Choose the way of running computations that fits your workflow; from always-on or on-demand services to large batch jobs.

Compute modeHostingNetwork access via campus network/VPNTypical workloads
24/7 APIsPersistent web endpointWith apps running on your machineReal-time inference, interactive data exploration, LLM gateways, agentic AI
24/7 Web AppsPersistent web applicationWith your browserInstant AI/ML web services and browser-based tools
On-Demand Web AppsSystem-configured scheduled jobWith your browserEfficient resource allocation for code development, testing and debugging
On-Demand Arbitrary ProcessingUser-configured scheduled jobWith your browser or a web-terminalAI/ML model training, complex research pipelines and massive batch inference runs
 

Compute Nodes

2,500 CPUs
  • 1,700 Intel CPUs
  • 800 AMD CPUs

Distributed Storage

1.5 PB
  • Network file systems (NFS)
  • S3-compatible buckets
  • Windows/macOS file shares

Schedulers

Slurm
  • Slurm batch scheduling
  • On-demand interactive apps, including MATLAB, QGIS and Stata

45 NVIDIA GPUs in two performance tiers: enterprise AI/ML hardware and mid-performance GPUs.

Tier 1

Enterprise AI/ML computing
  • 16× NVIDIA H10080 GB
  • 1× NVIDIA A100

Tier 2

Mid-performance
  • 4× NVIDIA RTX A6000 Ada – 48 GB
  • 10× NVIDIA V100 – 32 GB
  • 10× NVIDIA RTX 5000 – 24 GB
  • 4× NVIDIA RTX 4500 – 20 GB

Easy Access From Your Web Browser

You don't need command-line experience to get started. The ¸Ô±¾ÊÓƵ Supercomputer is available through Open OnDemand, a web portal widely used in academic research computing.

Easy Access From Your Web Browser

You don't need command-line experience to get started. The ¸Ô±¾ÊÓƵ Supercomputer is available through Open OnDemand, a web portal widely used in academic research computing.

  • Launch interactive apps such as RStudio, Jupyter notebooks, and Visual Studio Code
  • Upload and share data with a point-and-click file manager or tools like Cyberduck
  • Submit computing jobs with guided forms that write the job script for you
The Open OnDemand portal: manage files, open a web-terminal, submit jobs, and launch interactive apps from a browser.
The Open OnDemand portal: manage files, open a web-terminal, submit jobs, and launch interactive apps from a browser.
The guided Slurm job form for GPU compute nodes.
The guided Slurm job form for GPU compute nodes.

Request GPUs without writing a script

A guided form lets you request a GPU, up to 64 CPU cores, and up to 480 GB of memory on a GPU compute node. The Open OnDemand portal helps you write Slurm job scripts.

Ready-to-use scientific software

Researchers can create their own Conda and Python environments, or use pre-configured Conda and Python environments for PyTorch, TensorFlow, and other well-known frameworks that make deep learning and language analysis straightforward and reproducible.

Python, R, MATLAB, QGIS, Stata, AlphaFold, Gaussian, GROMACS, SageMath, Macaulay2, Magma, astroML, xtb

Private, On-Campus Artificial Intelligence

¸Ô±¾ÊÓƵ runs its own privately hosted open-weight large language models on university hardware. Prompts, documents and outputs stay on campus and are never sent to commercial AI providers. That matters for research confidentiality, student privacy, and academic integrity.

¸Ô±¾ÊÓƵ AI ()

An AI harness for research, teaching, and administrative work:

  • OpenAI-compatible LLM Endpoints accessible via API key
  • Chat with your own uploaded documents and web pages
  • Audio transcription and summarization
  • Speech-to-text, read-aloud and hands-free conversation mode
  • Temporary chats that aren't saved
  • Optional web search
  • Custom retrieval-augmented generation (RAG) knowledge bases for courses and projects
¸Ô±¾ÊÓƵ AI's chat interface, hosted entirely on campus.
¸Ô±¾ÊÓƵ AI's chat interface, hosted entirely on campus.

Research Applications and Web Services

Beyond raw computing power, ITS hosts specialized applications for research and teaching.

Deep Bryo segmenting colony structures in an electron microscope image.

Deep Bryo: AI for biology

DeepBryo is a machine-learning tool that automatically identifies and measures structures in scanning electron microscope images of bryozoans, small colonial marine invertebrates. ¸Ô±¾ÊÓƵ hosts it locally so lab members always have reliable access.

SealNet compares a new photo of a seal’s face against its library of known seals and suggests the five closest matches.

SealNet: AI face recognition for seals

SealNet is a custom AI-powered tool that helps track seal populations by automatically identifying known seal-faces in photos based on previously identified seal-faces in photos of seals in their natural habitat.

TranscriptionLabs uses AI to turn the speech in your audio and video files into written text, and can translate it into other languages.

Audio Transcription/Translation

TranscriptionLabs uses AI to turn the speech in your audio and video files into written text, and can translate it into other languages.

RStudio IDE

Run R in the browser with no installation.

JupyterLab

Isolated notebook environments for different use cases.

Label Studio

Annotate text, images and audio for machine-learning datasets.

GitLab

A self-hosted, on-campus code repository for private collaboration and continuous-integration pipelines.

Data Science Collaboratory

Interactive R/Shiny statistics apps at .

Custom Applications

If it runs in a container, ITS can deploy it.

Security, Privacy and Responsible Use

¸Ô±¾ÊÓƵ's research computing is built around careful data stewardship.

On-premises data

User data lives in secured, ¸Ô±¾ÊÓƵ-operated data centers.

Tiered access levels

General, Restricted (after training and approval) and Controlled (for Controlled Unclassified Information).

Compliance-minded policies

Aligned with frameworks such as NIST 800-171 and FERPA.

Disaster recovery

24-hour recovery time objective and 3-hour recovery point objective.

Published policies

Access and eligibility, usage, data management and storage, disaster recovery, and training and support.

Training and support

Orientation sessions, workshops, Kubernetes training, documentation and one-on-one consultations.

How to Get Started

The ¸Ô±¾ÊÓƵ Supercomputer is open to all active ¸Ô±¾ÊÓƵ faculty, students, and staff on the ¸Ô±¾ÊÓƵ campus network or connected through the ¸Ô±¾ÊÓƵ VPN. Getting started takes about five minutes.

  1. Off campus? Connect to the ¸Ô±¾ÊÓƵ VPN first.
  2. Follow the instructions to and sign-in to the Open OnDemand portal with your ¸Ô±¾ÊÓƵ account (SSO credentials).
  3. Wait about three minutes while your account and home directory are provisioned.
  4. Sign-in again and start computing.

Need shared project storage, a dedicated machine, a GPU priority allocation, an API token, or a custom application? Email rcd@colgate.edu for a consultation.

Research Computing Committee

The Research Computing Committee (RCC) serves as an advisory body tasked with supporting and evaluating research computing resources across campus. The primary goals of the RCC are to: monitor the research computing needs of the ¸Ô±¾ÊÓƵ faculty; advise the CIT, ITS, and the PDoF Office about research computing; and support communication about research computing across the campus community and with prospective faculty.

radhikari
Ramesh Adhikari
Associate Professor of Physics & Astronomy
(315) 228-6084
334 Ho Science Center
  • MWF  10:15am - 10:55am  (334 Ho Science Center)
  • R  9:55am - 10:55am  (334 Ho Science Center)
achianese
Anthony Chianese
Professor of Chemistry; Chair, Department of Chemistry
(315) 228-7718
305 Wynn Hall
  • TWF  1:20pm - 2:20pm  (305 Wynn Hall)
ndavray
Niranjan Davray
Chief Information Officer
ndiana
Nicholas Diana
Assistant Professor of Computer Science
(315) 228-7588
307 Bernstein Hall
  • MW  1:15pm - 2:45pm  (307 Bernstein Hall)
tdincer
Tolga Dincer
AI Research Software Engineer
jdonnelly1
Joe Donnelly
Associate Director of Discovery Services; Assistant Professor in University Libraries
(315) 228-4830
cducomb
Christian DuComb
Associate Dean of the Faculty for Faculty Recruitment and Development; Associate Professor of Theater
(315) 228-7993
105E McGregory Hall
mloranty
Mike Loranty
Professor of Geography and Environmental Studies; Chair, Department of Geography
(315) 228-6057
307 Ho Science Center
omccabe
Owen McCabe ’27
jstenberg
Joseph Stenberg
Assistant Professor of Philosophy
112 Hascall Hall
ktsoukalas
Kyriakos Tsoukalas
Associate Director of Research and High Performance Computing