Calgary Compute Lab

Calgary Compute Lab is a public research-and-build repo for local inference, agent systems, forecasting, speech/voice tooling, and practical frontier compute.

This is not a startup landing page. It is a working lab notebook for building, testing, and publishing applied AI systems from Calgary.

Why this exists

A lot of frontier AI conversation feels detached from place. Calgary Compute Lab takes the opposite approach:

• build with local hardware
• document real experiments
• publish results in public
• connect research ideas to practical tools
• make Calgary feel like part of the frontier, not downstream from it

Research themes

1. Local Inference Systems

Running useful models on local and edge hardware, with an emphasis on cost control, reproducibility, and practical deployment.

2. Agent Infrastructure

Multi-agent workflows, orchestration patterns, tool use, memory systems, and human-in-the-loop operations.

3. Speech and Voice Systems

Text-to-speech, transcription, voice cloning, low-latency audio pipelines, and local GPU speech inference.

4. Forecasting and Decision Systems

Applied forecasting, scenario modeling, uncertainty-aware workflows, and decision support for builders, operators, and small teams.

5. Public Compute Culture

How cities like Calgary can build stronger technical communities around open tools, shared infrastructure, and visible experimentation.

Current directions

• Intel Arc B580 local GPU experiments
• Qwen3-TTS local voice cloning workflows
• OpenClaw agent infrastructure and ACP patterns
• local-first AI operations
• community-layer tooling for Calgary developers and builders

Philosophy

We care about systems that are:

• technically real
• locally runnable
• publicly explainable
• useful beyond a demo
• interesting enough to attract serious builders

Repository structure

notes/         Public lab notes and experiment logs
experiments/   Reproducible experiments and prototypes
docs/          Concepts, methods, and system overviews

First principle

If a model, workflow, or agent stack cannot be explained clearly, tested locally, and improved in public, it is probably not ready.

Calgary angle

This repo is intentionally place-based. Not because Calgary needs a gimmick, but because technical ecosystems become stronger when people can point to real local work.

Status

Active. Early. Public by design.