Canonical: https://northloom.ai/articles/

ArticlesPrivate AI / Practical decisions

# Make AI work  
for your business.

What to test, what to measure and what to ask before committing to private AI. Guides for the people planning, buying and running it.

October 2026 / Getting started

## [Where should your business start with AI?](https://northloom.ai/articles/where-should-your-business-start-with-ai)

Choose a useful first AI project by comparing the work your team already does. A practical example takes you from three ideas to a clear, one-page brief.

[Read the guide ↗](https://northloom.ai/articles/where-should-your-business-start-with-ai)

October 2026 / Model suitability

## [Can a local open-weight model replace a hosted AI assistant for company work?](https://northloom.ai/articles/can-a-local-model-replace-a-hosted-ai-assistant)

For some internal workflows, yes. The decision depends on your documents, error tolerance and deployment, rather than whether a local model wins a public leaderboard.

[Read the guide ↗](https://northloom.ai/articles/can-a-local-model-replace-a-hosted-ai-assistant)

October 2026 / Inference engineering

## [How to run large Mixture-of-Experts (MoE) models on limited VRAM](https://northloom.ai/articles/how-to-run-large-moe-models-on-limited-vram)

A practical guide to expert offloading, PCIe bandwidth, system RAM and context memory, informed by Northloom’s Rust and CUDA inference engineering.

[Read the guide ↗](https://northloom.ai/articles/how-to-run-large-moe-models-on-limited-vram)

October 2026 / Deployment costs

## [Local LLM or cloud API: how to compare the real costs](https://northloom.ai/articles/local-llm-or-cloud-api-cost-comparison)

Compare private AI hardware with a cloud API using workload volume, useful output, power and operating time. A worked example shows where the financial decision changes.

[Read the guide ↗](https://northloom.ai/articles/local-llm-or-cloud-api-cost-comparison)

October 2026 / GPU memory

## [How much GPU memory does a local LLM actually need?](https://northloom.ai/articles/how-much-gpu-memory-does-a-local-llm-need)

Estimate model weights, context memory and concurrent requests before buying a GPU. Worked examples explain why a model file that fits in VRAM can still run out of memory.

[Read the guide ↗](https://northloom.ai/articles/how-much-gpu-memory-does-a-local-llm-need)

October 2026 / Runtime selection

## [Choosing a local inference runtime: llama.cpp, Ollama, Krasis and the alternatives](https://northloom.ai/articles/local-inference-runtime-comparison)

A practical comparison of local AI runtimes, from desktop tools to shared inference servers and specialist engines for models larger than GPU memory.

[Read the guide ↗](https://northloom.ai/articles/local-inference-runtime-comparison)

October 2026 / Inference performance

## [Why your local LLM is slow, and how to identify the bottleneck](https://northloom.ai/articles/why-your-local-llm-is-slow)

A long wait before the first word and a slow stream of words are different problems. Separate them, measure a repeatable workload and investigate the stage that is taking the time.

[Read the guide ↗](https://northloom.ai/articles/why-your-local-llm-is-slow)

October 2026 / Model suitability

## [Can a local model actually handle your company's workload?](https://northloom.ai/articles/can-a-local-model-handle-your-workload)

A convincing demonstration is a useful start. A small, structured evaluation tells you much more about whether a local model is ready for everyday work.

[Read the guide ↗](https://northloom.ai/articles/can-a-local-model-handle-your-workload)

October 2026 / Infrastructure planning

## [Before buying an AI server: what to measure and ask your supplier](https://northloom.ai/articles/before-buying-an-ai-server)

An AI server is only a good purchase if it can do your work at an acceptable speed. Start with the workload, then ask for evidence that the proposed system can handle it.

[Read the guide ↗](https://northloom.ai/articles/before-buying-an-ai-server)

AI Consultancy

## A decision to make?  
Bring us the details.

Independent engineering advice on your proposed hardware, model suitability or inference performance. Defined assessments with written findings.

[Explore services & pricing ↗](https://northloom.ai/consultancy/)
