Equations Work: Intelligence,
engineered.
Building intelligent systems since 2011.
You are inside the black box.
We have worked here since 2011.
Room I — Inside the Black Box
black boxnoun
- A system whose inputs and outputs can be seen, but whose inner workings cannot.
- How most organisations meet artificial intelligence: a question goes in, an answer comes out, and what happens in between is taken on trust.
Inside the box is where we work: the data, the models, the reasoning and the checks between the question and the answer.
Deep tech is not our pivot.
It is where we started.
Equations Work has been building intelligent systems since 2011 — through statistical learning, deep learning, the transformer, foundation models and now agents. Each era brought new models. None of them removed the engineering that makes intelligence dependable. That continuity is what we bring to every system we build.
- 2011 – 2013Learning from featuresIntelligence meant hand-engineered features, statistical models and rules — narrow, brittle, and only as good as the data pipeline beneath them.It demandedData engineering, domain modelling, and evaluation against the real world rather than the test set.
- 2014 – 2016Deep learning reaches productionAfter the 2012 ImageNet result, neural networks overtook hand-built features in vision and speech. Accuracy jumped — and so did the appetite for data and compute.It demandedGPU infrastructure, training pipelines, and serving models within real latency budgets.
- 2017 – 2019Attention changes languageThe transformer (2017) and pretrained language models made text a first-class input. Understanding documents became an engineering problem, not only a research one.It demandedRetrieval, document understanding, and adapting general models to specific domains.
- 2020 – 2022Foundation modelsVery large pretrained models began to generalise across tasks. Capability arrived as an API call. Reliability did not.It demandedEvaluation, grounding, guardrails, and choosing the right model for the cost and the risk.
- 2023 – nowAgents and systemsModels now plan, call tools and act across systems. The hard problem has moved from the model to everything around it.It demandsOrchestration, state, validation, governance and human oversight — the subject of the rooms that follow.
One unbroken line through five eras — Equations Work, 2011 to the present.
Five eras. The models changed in every one of them. The questions did not: where does the intelligence sit, what may it touch, how do we know it works, and what happens when it fails?
See the disciplines that answer themRoom II — The Systems
Six disciplines, nested around the problem.
Not a stack every product shares. A set of disciplines every working system draws on — in different proportions, for different problems. Read from the surface inward.
Room III — The Inventions
Five inventions.
Each one began as a problem without machinery.
Some problems need a product. Others need the machinery beneath it invented first. Look closely at each work — every painting changes when inspected — then open it to examine the mechanism.
Our inventions also become components: the same machinery appears inside enterprise systems we engineer for clients.
Room IV — The Engineering Practice
Not every problem should become a product.
We engineer systems that must live inside existing architecture, policy and operations. Sometimes that means deploying one of our inventions. Often it means combining proprietary components, bespoke engineering and applied research into a system shaped to the problem — not the other way round.
What we take on.
The engagements differ. The discipline does not. Our proprietary technology strengthens this work; this work exposes the problems that shape our research.
Room V — Evidence of Work
Evidence of work.
Domains, not names.
Client names stay with our clients. What remains is the engineering: the problem, what we built, the idea that made it work and what changed. Three measured outcomes below. The full register of systems we have delivered across industries is in the Cabinet.
Every record is anonymised by design. Ask about a comparable problem and we will tell you what we can.
Room VI — The Open Problem
Bring us the problem that doesn’t fit the template.
If it sits between models, software, machines and people — and nobody has a ready answer — describe it. A few sentences are enough to begin. The square is yours to fill.