Optimisation API development
Your product has to decide on its own, in real time, with no planner to check. We build the optimisation engine it calls.
- Free 30-minute call
- No commitment
- You own the code
In short: An optimisation API is an interface that lets software request a decision, such as a route or a schedule, and get the computed plan back. It runs as a service with a fixed time limit, so a product can decide without a planner.
The problem
A customer books a delivery slot in your app. Your product has to decide on the spot which van takes it, and nobody is there to check. Today a simple rule decides: the first slot that fits wins. It answers fast, but nobody knows how good the answer is.
What changes
Today, a fixed rule makes each decision, and nobody sees what it costs. With our solution, your product asks an optimisation engine for every decision. It gets the best answer found within the time you allow. Longer jobs, such as tomorrow’s full plan, run in the background without slowing the app. Your engineers can call the engine, monitor it and trust it.
How it works
We design optimisation services as APIs, interfaces your software calls like any other service. REST endpoints, the web addresses your app sends requests to, handle decisions that need an instant answer. Async job queues take longer-running models: the app hands over a job and collects the result later. Every service has an explicit latency budget, the longest it may take to answer.
We use anytime algorithms, which return the best solution found within your time limit. LeagueOptima, our fixture-planning product for sports leagues, shows that we build and run optimisation software in production. We apply the architecture behind that live product to your system.
The questions your engineers will ask
First, the licence. Many teams run their software in containers, sealed packages that run the same on every server. Docker and Kubernetes are tools for that. A commercial solver inside a container needs a licence built for it. Gurobi’s Web License Service is one, and Gurobi states that the containers must reach the internet to use it. Open-source solvers skip that step: HiGHS, for one, comes under the MIT licence. We settle this before the architecture, because it decides where the service can run.
Second, the clock. The OR-Tools documentation recommends a time limit on every CP-SAT solve, and we set one on every engine. When time runs out, the service answers with the best plan so far. It also reports the status and the gap: how far that plan may still be from the best possible one. Your app can then decide whether to accept it or fall back to the old rule. The honest boundary: if you make the decision once a week, you don’t need an API. A scheduled model run is simpler and cheaper.
When it isn’t the answer
The honest boundary: if you make the decision once a week, you don’t need an API. A scheduled model run is simpler and cheaper.
Common questions.
Something else? Ask us directly.
Yes, with the right licence. Gurobi’s Web License Service is built for containers and cloud platforms. It issues tokens that renew automatically, and Gurobi notes that the containers need internet access for it. If your service must run without internet, that needs a different setup, so we clarify it early.
The service still answers on time. It returns the best plan found so far and a status. It also reports how far that plan may still be from the best possible one. If there’s no usable plan yet, it says so, and your app falls back to its old rule.
Ask whether a person is waiting for the answer. A customer picking a delivery slot is waiting, so that request gets its answer within a fixed time. Nobody is waiting for tomorrow’s full route plan, so it goes into a job queue and your app collects the result later. Many products need both, sharing one engine.
By logging it like any other service. We record inputs, solve times and how far each plan may still be from the best possible. We also count how often the fallback rule steps in. Practitioners on Operations Research Stack Exchange add one more signal: users who re-run the optimisation again and again don’t trust its plans. When those numbers get worse, you see it before your customers do.
Sometimes they can, and then we’ll say so. Standard tools are built for an average problem, though. With Optimisation API development, we start from yours: its size, its rules and the exceptions nobody wrote down. That is why the plan fits your operation better.
You don’t need either to start. We build the model from the data you’ve got, messy or not, and cleaning it up is part of the job. You get a documented system your team can maintain, and we train them until they can run it long after we’ve left.
It starts with a free 30-minute call about your operation and the model that would fit it. Then we build a working prototype and run it on your numbers. You’ll see results before you commit to anything bigger.
Similar pages.
Terms on this page
Every abbreviation explained in plain words, in our article “Metrics and methods”.
Tell us what you’re planning.
In a free 30-minute call with a specialist, you find out whether optimisation can help.
- Free, no commitment
- You talk to a specialist directly
- A clear go or no-go