Compute
In BC Gov, compute means enabling diverse models and applications across on-prem and public cloud hardware. Most current research focuses on measuring on-prem CPU and GPU performance.
Program
Research
Our team investigates compute, model internals, agentic delivery, and safety practices so AI can be applied responsibly and effectively across BC Government services.
Why this research
- Find cost-effective methods to use and deploy AI in BC Gov services.
- Understand AI behavior deeply to make safer implementation choices.
- Choose the right AI approach for each application and service context.
- Track trends and fast model changes over time.
- Improve dataset preparation, training methods, and compute choices for safe AI for citizens.
Available now: Compute Agentic SDLC
Scope at a Glance
Four research pillars + enablement
AI Internals
Models and tooling change quickly at the surface. Deeper understanding of model behavior and failure patterns helps us build AI with our own architecture choices and datasets over time, including safer dataset preparation and training methods.
Applications and Agentic Systems
We study agentic systems, tool orchestration, and human-in-the-loop controls so workflows are reliable in production, not just in demos. The first write-up is what this team learned running remediation and greenfield rebuilds.
Safe AI and Red Teaming
Threat tests, red teaming, eval suites, and rollout safeguards are treated as core engineering requirements for public-sector use.
Enabling AI Across BC Gov
Reusable implementation playbooks for policy, operations, and delivery teams across ministries.
Publications
| Date | Title | Stream | Type |
|---|
No publications match this filter.