Data Science & Models
Unlocking the Value Hidden in Your Data
Gathering quality data is only the starting point. It is the clear interpretation that turns data into insights which can truly help your team and your business. Here at Funis Consulting, we build and use custom advanced analytical and intuitive visualisation methods to transform raw information into actionable narratives supporting you in daily or ad-hoc business decisions. Whether it is a dataset which spans years or a dataset which captures a moment in time, we help you use your dataset to its fullest potential, so that you can understand it, communicate it, and most importantly act on it.
-- Visualisation of complex data to determine trends easily --
Your data can also be used to build models to predict a specific scenario’s outcome. These can range from basic statistical models such as linear and non-linear regression to machine learning models including neural networks, random forests, off-the-shelf or custom clustering algorithms, and AI-driven solutions.
We excel in pattern recognition for multi-dimensional datasets and, through dimensionality reduction and custom algorithms, we have experience in uncovering hidden relationships between datapoints. This helps in exposing trends and avoiding misleading classifications. A significant part of our work also involves forecasting future system states using models calibrated with real historical data.
-- Clustering datapoints using custom clustering algorithms --
We know how flawed or incomplete data leads to skewed predictions, and hence poor decision-making, and so, gathering quality data across multiple channels is of utmost importance. Funis Consulting can assist you in designing the best data gathering strategy so that you have good quality data to use for understanding your system or building and training your models.
All of this is done by building tailored software and solutions that allow your organisation to analyse, visualise and interpret complex datasets with ease, or predict with a degree of confidence what a future state will look like, helping you with your strategy and decision-making.
First-Principles Modelling & Simulation
Mathematical, Chemical and Physical Models
Data is incredibly powerful and so when you use data-driven models, complete with pattern and trend prediction, you can go far when it comes to understanding your audience or your system. Having said that, data-driven models, as powerful as they may be, have their limitations and that is especially obvious when it comes to predicting behaviour far beyond the range of the data that is available. Extrapolating too far introduces a lot of uncertainty and can quite easily lead to incorrect predictions, with very wrong decisions being taken as a result.
So, what do we do in such cases? Many real-world processes can be modelled directly from first principles, such as laws of physics or complex logic. These types of models offer far greater flexibility enabling you to explore new operating conditions, diagnose complex behaviours and gain a deeper insight into how your system truly works.
.png)
Which way to go with depends on what you have or know. If you have lots of quality data, but not as much knowledge, then data-driven models are best to use. Alternatively, if the dataset is not as large, but you do know how your system works and can describe that using an equation or a set of equations, however complex, then first-principles models are best suited for you. And if you have both the data and the knowledge then you can go with either data-driven models of first-principles, or else get the best of both worlds and combine the two methods together.
-- Molecular Dynamics --
Our expertise, both in-house and through trusted collaborators, covers a wide range of advanced first-principles modelling techniques. These include:
-
Reaction Kinetics and Thermodynamic modelling
-
Molecular Dynamics (MD)
-
Process Modelling
-
Computational Fluid Dynamics (CFD)
-
Finite Element Analysis (FEA)
-
Discrete Element Modelling (DEM)
-
Discrete Event Simulation (DES)
These methods are not limited solely to science or engineering applications. In fact, we have successfully applied physical-modelling principles to economic and operational systems bringing clarity to environments where intuition and experience alone may fall short.
-- Output from a Reaction Kinetics and Thermodynamics Predictive Model --
We also have experience in building fast surrogate models by combining data science with first- principles modelling training machine-learning algorithms on Monte Carlo outputs from complex physics-based models, reducing prediction times from minutes or hours to near-instant. This is fundamental for speeding up decision-taking through instantaneous predictions.
In summary, we can support businesses and teams across any sector in building models. These can range from simple chemical reactions to highly complex multi-physics processes. Through equation-based modelling & simulation as well as a problem-solving focused mindset, our solutions are tailored to your needs.
Process & Parameter Optimisation
Achieving Desired Outcomes by Determining the Best Set of Parameters
Optimisation applies to all industries. If you know what you would like to achieve but are unsure of which parameters to change, how and when, that is where Funis Consulting can help you. Across sectors, manufacturers and operators face variations of the same challenges, whether that’s tight margins, high energy costs, supply chain uncertainty, labour shortages, tight timelines, or constant pressure for quality and consistency. Despite widespread automation, many operations still rely on reactive decisions and trial-and-error adjustments. Valuable experience remains essential, but in today's fast changing world the tools exist to move from intuition to informed decision-taking.
-- Prediction versus Optimisation diagram --
Any optimisation tool starts with a robust predictive model, whether that’s data-driven or based on first-principles. With such models you can test ideas in a virtual environment before changing anything in real life. Digital models mimic the behaviour of your process or system, allowing you to explore “what-if” scenarios, identify constraints, forecast outcomes and design improvements safely and efficiently.
Optimisation goes a step further and finds what are the optimum input parameters for your system so that you can achieve the desired outcome. This can be done either in two ways. The first is to carry out hundreds or thousands of “what-if” simulations and then analysing all of the resulting data. This can be used to not only determine the optimal parameter values, but also determine the sensitivity of your system to specific inputs. The second is to use machine-learning algorithms, such as gradient descent or Bayesian techniques, to quickly home onto the optimal parameter settings.
-- Rapid convergence using Bayesian Optimisation --
Optimisation can be single-objective or multi-objective, meaning you can optimise your parameters for a single outcome (e.g. lowest cost) or multi-objective (e.g. lowest cost and shortest time). Multi-objective optimisation can be tricky, as the objectives can be conflicting (e.g. a lower cost means a longer production time and vice-versa). Multiple models can also be linked to capture interactions across processes. This is where Funis Consulting comes in, advising on how to approach such optimisation problems and testing out various possibilities before communicating all to you to find a solution that works best for you.
This work requires more than code. It demands a deep understanding of how physical and chemical processes behave, and how to translate that knowledge into models that work in practice. In fact, we work alongside domain expertise for most of such projects as we believe that a joint effort is required to fully understand a system in a holistic way.
Human-Centric Solutions
Solutions Designed and Built with and for People
Here at Funis Consulting Ltd, we firmly believe that technology and innovation should serve a purpose. Through our work we create impactful solutions that do good... solutions that make a positive impact. We focus on delivering smart, tailored solutions that create real value, that empower rather than overwhelm, designed and built with the end-user in mind, whatever the level of technical ability of the user. This ensures that those stakeholders who need to use the solution or need to understand the insights produced from it can use the tool with confidence and without difficulties, whether these are operators, shop-floor personnel, managers or data analysts. Our interfaces are designed with different views, access rights and visual styles. This is because a data scientist and a machine operator at your company may not need the same level of detail and so they do not need the same interface or outputs. Or if they do need the same details, the interface can be made to display differently according to the team's needs on the job as well as the level of technical, scientific and data proficiency of the users.
Our approach is inclusive by design, and we use clean layouts, accessible language and thoughtful UX choices, making the software or solution comfortable to use as well as quicker and easier to adopt within the business and amongst teams. We ensure that we listen to the needs of business as well as of the different team members utilising the software, at different stages of development and we work alongside in-house personnel during the software development design, wherever this is possible, so that their input, concerns and preferences are taken on board at an early stage. This ensures true customisation and usefulness of the tool.
-- User-friendly UI built for FuniSoft Ltd --
Beyond design, we also support the people who will use the solution. We have experience in training and capacity-building and we can provide onboarding, tailored training sessions and early-stage guidance so teams feel confident in adopting the tool with ease and in a short period of time.
Even though our solutions may rely on science, mathematics, data and state-of-the-art technology, such models and algorithms are running and working for you in the background, while what teams interact with is a clear and intuitive interface or dashboard designed around the user, therefore making the experience for your business and teams straightforward, purposeful and human. While Python is our coding language of choice due to the versatility and vast libraries and possibilities in the scientific realm, we are software-agnostic, meaning that we can also work with other programming languages and software, ensuring seamless integrations into your current systems.
Ultimately, Funis Consulting’s goal is to deliver solutions that fit naturally into real workflows, simplifying decision-making and empowering users. We do not simply build technology, but we focus on enabling clarity, usability and thoughtful design with our solutions, for the end-goal of creating real positive impactful solutions that do good.