Home / Lab / Toolkit

The toolkit we bring to your project.

Every tool here started as something we needed in our own lab. Each one maps to the product or service that uses it.

ClusterSpace

One tiled workspace for terminals, SSH, browsers and AI agents. Goals that finish, and a dashboard for the whole fleet.

OpsAgentsMIT

Used in: Hardware · Implementation

GPU & fleet ops

nvidiacp

GPU tuning and control for the boxes we build and run.

GPUTuning

Used in: Hardware

tempmon

Thermal watchdog that keeps GPUs inside safe limits around the clock.

GPUMonitoring

Used in: Hardware

MuxTer

See into every terminal session across the fleet at once. ClusterSpace's companion.

OpsTerminals

Used in: Hardware · Lab

Inference & eval

vllm-topology-bench

Benchmarks how to split models across GPUs. Showed that two smaller replicas can beat one big tensor-parallel split on the same cards.

vLLMBenchmark

Used in: The Engine · Hardware

model-chat-cli

Blind-test models side by side on your own prompts before you pick one.

EvalCLI

Used in: Consulting

Agents & automation

graphx

Graph pipelines with human approval gates and connectors.

PipelinesHuman-in-the-loop

Used in: Implementation

cadillac-builds

Apps built end to end by cadillac with zero human edits, kept as a public record.

Proof

Used in: Implementation

Model R&D

tokopt

Tokenizer adaptation: fit a model's vocabulary to your domain.

TrainingTokenizers

Used in: Implementation

REAP-AWQ research

Pruning plus 4-bit quantization of a large coding model so it fits on hardware a business can own.

QuantizationPruning

Used in: Hardware

Dev environment

code-atlas

Map a codebase in 3D. We use it to audit a client's code before we touch it.

AuditVisualization

Used in: Consulting

cliide

A terminal-first IDE for working next to agents.

TerminalIDE

Used in: Lab

Start with a free exploration call.

Tell us what you're trying to do and what's in the way. If AI fits, we'll show you where. If it doesn't, we'll say that too.