← All services
Scientific AI
HPCg helps IOCB researchers put AI to work in their science. We run AI models locally on Aurum, so sensitive
data stays in-house, and support projects from the first idea to validated results.
Services
Local Inference Service Testing
Use AI models hosted on Aurum through browser-based chat or an API for scripts and applications.
- On-site
- processed on Aurum GPUs
- No training
- on your prompts or files
- Not stored
- by the model backend
- Up to 512k
- tokens of context
Current models:
Qwen3.8 27B
Qwen3.8 Flash Next testing
MiniMax-M3
GLM-5.3
GLM-5.3 Flash
Compare models →
LLM chat
Chat in your browser with tools for research and documentation:
File Search for PDFs
PubMed
Aurum docs and issues
Agents
Web search opt-in
Chat features →
LLM chat keeps your chat history, which you can delete at any time. For sensitive data, prefer the API and leave web search off.
Ask HPCg
Ask questions about the HPCg and AI documentation, including the IOCB AI policy. No login; answers link to their sources.
Ask HPCg → (IOCB network or VPN)
AI coding assistants
Use the Aurum models in OpenCode, in the terminal or in VS Code. Coding agents can run commands, so run them isolated.
Set up OpenCode → Isolation guide →
Document search for your group
A private, searchable collection of your group's PDFs, available in LLM chat. Set up on request.
Contact us →
AI solutions and workflow integration
Design and integrate AI tools, MCP servers, models, and data sources into scientific workflows.
Scientific AI project consulting
Assess feasibility, choose methods, identify data and computing needs, and plan scientific validation.
Advanced AI training
Domain-specific training on local models, workflow integration, evaluation, and reproducibility.
Personnel
Scientific AI Lead
Also serves as IOCB AI Ambassador and AI Project Manager as defined in the IOCB AI policy.
AI Infrastructure Engineer (Aurum)
Scientific Imaging and AI Specialist
What we cover
| Area | What it includes |
|---|
| Scientific AI leadership and governance | Setting the direction for scientific AI and connecting research needs, technical delivery, policy, and institutional coordination. |
|---|
| AI solutions architecture and engineering | Designing, building, and integrating AI tools, MCP servers, models, and data sources into scientific workflows. |
|---|
| Local AI infrastructure and platform engineering | Designing, deploying, and operating local AI infrastructure on the Aurum cluster and the services hosted on it. |
|---|
| Scientific AI project planning and consulting | Helping researchers define projects, assess feasibility, select methods, identify data and computing requirements, set validation criteria, and plan implementation. |
|---|
| Advanced scientific AI training | Developing advanced and domain-specific training linked to scientific tools and services. |
|---|
Examples of our work
- Developed IOCB's AI policy and practical guidance for secure, responsible, and scientifically sound AI use.
- Built MCP servers and integrated AI with approved scientific software, services, data sources, and research workflows.
- Deployed locally hosted AI models and inference services on Aurum and made them available to IOCB researchers.
- Helped research groups define feasible AI projects, select appropriate methods, and establish scientific validation criteria.
- Delivered advanced and domain-specific training on local models, workflow integration, MCP servers, evaluation, and reproducibility.
Scope
HPCg supports advanced and scientific uses of AI. We do not provide general AI assistant support or routine end-user troubleshooting. ITS provides the mandatory introductory AI training; HPCg provides advanced and domain-specific training and coordinates with ITS where scientific AI services depend on shared institutional systems.
