Independent research lab · St Andrews, Scotland

Research that
compiles.

We turn hard questions in AI, systems, optimisation, and privacy into open, testable software—so the work can be inspected, challenged, and built on.

Explore the work Work with us
RESEARCH LOOP / 01 OPEN
Hypotheses → runnable artefacts Public by default
18 public systems
05 research domains
06 implementation languages
SC809174 registered in Scotland

Research areas / 01

Five connected fields.
One applied practice.

We work where machine intelligence meets the systems it depends on—from agent protocols and retrieval to runtimes, optimisation, and privacy.

01 LLM & Agents Prompts, agent protocols, routing, and client-side LLM plumbing. 06 systems 02 Search & Retrieval Embeddings, vector search, answer engines, and deliberative search. 04 systems 03 Systems & Runtime Sandboxing, NUMA-aware runtimes, and correctness oracles. 03 systems 04 Optimisation & Decision Constraint solving, ranking, trading signals, and simulation. 03 systems 05 Privacy & Trust Metadata-private messaging, key hygiene, and SQL data protection. 02 systems

Selected systems / 02

Ideas made
inspectable.

Each project is a runnable research artefact: source, assumptions, and implementation open for scrutiny.

View all 18 systems
01 Rust

Agent protocol

MPL

Your agents call tools and each other and you need contracts, scoring, and provenance.

Explore system
02 Rust

NUMA runtime

NumaPerf

You are hunting p99 tail latency on multi-socket NUMA hardware.

Explore system
03 Rust

Kernel correctness

GPUEmu

A single torch.allclose is not enough to trust your custom kernels.

Explore system
04 Rust

Robotics simulation

WareMax

You need a deterministic RMFS dispatching benchmark for RL or policy work.

Explore system
05 Python

NL to solver

Savanty

You have a discrete constraint problem (scheduling, assignment) to state in plain English.

Explore system
06 JavaScript

Prompt DSL

Promptel

You want prompts as typed, provider-portable, versionable artefacts.

Explore system

Operating model / 03

Hypotheses
as software.

A paper explains an idea. A working artefact exposes it to reality. We use both, but build early enough for implementation to challenge the theory.

Read our research approach
  1. 01

    Frame the question

    Start with a narrow claim that can be disproved, not a broad promise that cannot be measured.

  2. 02

    Build the instrument

    Encode the hypothesis in software with explicit interfaces, constraints, and failure modes.

  3. 03

    Test the edges

    Probe correctness, behaviour, and trade-offs under conditions that make assumptions visible.

  4. 04

    Publish the artefact

    Release the implementation so others can inspect it, reproduce it, and extend the work.

Collaborate / 04

The useful work is
rarely solitary.

We are looking for partners with meaningful questions, complementary expertise, or a setting where open research can create practical value.

Explore partnerships admin@skelfresearch.com
01

Academic

Joint research, validation, supervision, and reproducible artefacts.

02

Public sector

Applied R&D, trustworthy systems, and evidence-led technical work.

03

Industry

Research collaborations, technical partnerships, and focused prototypes.

04

Open source

Maintainers, contributors, adopters, and organisations building in the open.

Frequently asked questions about Skelf Research

What is Skelf Research?

Skelf Research is an independent AI research laboratory based in the United Kingdom. We investigate machine reasoning, retrieval, systems performance, and privacy, and publish inspectable software. Our current portfolio is a family of 18 public products spanning LLM & agents, search & retrieval, systems & runtime, optimisation & decision, and privacy & trust.

What does Skelf Research build?

18 public products, including promptel, blogus, mpl, route-switch, anouk, direktor, embedcache, memista, polymathy, slorg, zviz, numaperf, gpuemu, savanty, compere, waremax, l0l1, and perishable. 17 currently carry a detected MIT, Apache-2.0, or GPL-3.0 licence; Perishable's source is public while its licence is resolved.

Where can I find each product?

Every product lives at its own subdomain in the form <name>.skelfresearch.com — for example promptel.skelfresearch.com or zviz.skelfresearch.com. Documentation is at docs.skelfresearch.com/<name>/ and source is at github.com/Skelf-Research/<name>.

Is Skelf Research open source?

Open source is central to our methodology. 17 of the 18 current public product repositories carry a detected MIT, Apache-2.0, or GPL-3.0 licence. Perishable is public-source but is not described as open source until its licence is added.

Where is Skelf Research based?

Skelf Research is based in the United Kingdom (registered in Scotland, company no. SC809174). We work with academic collaborators, research partners, and funders globally.

How do I choose which Skelf product to use?

Start from the problem. If you need typed, portable prompts use promptel; to lock prompts in CI use blogus; for agent-to-agent contracts and audit use mpl; for local embeddings use embedcache; for vector search use memista; for a container sandbox use zviz; and for natural-language optimisation use savanty. The products page groups all 18 current systems by research domain.