I am finishing my studies at the Singapore University of Technology and Design under the pillar “Engineering Systems and Design”. I still struggle to explain what exactly my pillar offers.

It is easy to understand what other pillars in SUTD does

  • Architecture and Sustainable Design (ASD) - architecture
  • Engineering Product Development (EPD) - build physical stuff
  • Information Systems Technology and Design (ISTD) - build computer stuff
  • Design and Artificial Intelligence (DAI) (new pillar) - build stuff with a focus on AI

The pillar I am in is Engineering Systems and Design (ESD).

Pillar descriptions

We first look at some description of the ESD pillar.

This is a description by the Head of Pillar Peter Jackson.

Engineering Systems and Design deals broadly with large-scale systems and the design, the planning and the management of these systems.

This is an overview of the ESD pillar posted on the school website.

How do you decide what company to invest in, where do you locate that new factory, how can you be both ‘green’ and efficient, when should you launch that next generation product, whom should you target with your new internet service, and why might some application not be the best use of a new technology? These are invariably systems decisions, with many complexities and uncertainties. Design, analysis, and optimisation are how you tackle these open-ended problems.

This is how I introduce myself and my course of study in my technical interviews.

I am a final year student from Singapore, studying at Singapore University of Technology and Design. In my course, Engineering Systems and Design, we use math and code to model company processes to analyse and optimise them. I participate in coding competitions, on Codeforces I am a candidate master, on Kaggle I am a notebooks expert.

In a recent dialogue with the ESD Head of Pillar Peter Jackson, I ask him how would he introduce ESD in a job interview. This is his response.

Collecting together the engineering disciplines that deal with systems, that is the core of ESD. In other universities, these areas are split into different domains: Management, Industrial management, Chemical engineering, etc. Here in ESD we pull together the core elements in these domains from these multiple domains and put them together, blending data science, decision science, finance and economics, logistics and process optimization. This is what ESD is about.

Courses studied in ESD

A main part of university education is the set of courses taken. This is the list of core courses all ESD students will need to take

  • Optimisation. We study the Simplex algorithm to solve linear optimisation problems, algorithms to solve the network flow problems. We are also exposed to some concepts from dynamic programming and integer programming. Optimisation is about maximising or minimising an objective subject to the constraints, and in this course, the objective and constraints are linear.
  • Data and Business Analytics. One half of the course teaches basic finance and accounting concepts, and the other half of the course teaches basic data analysis tools. The school would usually source projects from companies for students in assigned groups to do.
  • Probability and Statistics. I took these courses in SMU instead of SUTD, I could not comment on how this course is conducted in SUTD.
  • Manufacturing and Service Operations. Manufacturing involves managing inventory level, and the question is how much inventory to manage and how much inventory should you refill, and these courses explores the optimal solutions given the assumptions. Service involves queues, and we understand the performance of a modelled queue. The group project requires us to work on a manufacturing operations or service operations problem from a company that we need to find.
  • Engineering Systems Architecture. We were also taught some tools for model our solution as states and subsystems, and well as some concepts of multi-objective optimization. The project requires us to use these decision tools to build a game.
  • Simulation, Modelling and Analysis. We were taught tools to design simulations, and use statistics to estimate the performance of the simulations. The group project requires us to build a simulation and analyse the findings.

The following are electives that ESD students can take. With a sufficient combination of courses, the track can be recognised in the transcript.

  • Supply Chain - Supply Chain Management and Supply Chain Digitalisation and Design
  • Business Analytics and Operations Research - Statistical Machine Learning, Advanced Optimisation and Stochastic Modelling, Networked Life, Game Theory
  • Finance - Investment Science, Equity Valuation and Financial Systems Design
  • Aviation - Airport Systems Planning and Design, Airport System Modelling and Simulation

The value of ESD studies

Business degrees, like the one I did in SMU, has half of the courses being more quantitative (math and formulas) and half of the courses qualitative (concepts and frameworks). Usually, quantitative courses have projects that are more quantitative in nature, and therefore the content is more like two-thirds quantitative, one-third qualitative.

ESD can be viewed as a business degree from the perspective of engineering - which is more than 90% quantitative. We do have our dose of qualitative courses from humanities courses that we need to take almost every term.

ESD and business degrees provide us methods to help us to make better quantitative decisions. In a business, the decisions can also be classified as quantitative and qualitative. Qualitative decisions are usually more important as it concerns the entire strategy of the company and decide which problem do we even solve. When the problem direction is define, the quantitative decisions come in - some of these decisions include location, price and performance.

The quantitative decisions need to be backed with data and assumptions, which is what is being done in ESD.

To achieve any of these, we usually need to identify the following elements in our solution

  • Task. What are we trying to improve? For example, we want to improve the waiting time between buses.
  • Objective function. How do we measure that we have improved? We need to calculate the expected waiting time, and we can use simulation to measure the improvement in waiting time.
  • Data. We need to know that our improvement is grounded in reality. What data do we have to support our modelling? We want to ensure that our simulation makes sense. We will compare our simulations with the data collected.

These are the actionable suggestions that the ESD models can provide, and we can define and measure the performance of these suggestions with data

  • Prediction. Given past data and current information, what is the prediction of a certain value? In the same example, if we increase the bus frequency, how much the wait time would decrease?
  • Prescription. Given the past data and current information, what actions do we recommend for the decision-maker? Again in the same example, we can suggest the optimal bus frequency for bus operator, considering both the satisfaction of the rider and the increased operating cost.
  • Policy. Given the motivations and characteristics of each decision-maker, how can we advise the policy-maker to set up the system to achieve desirable outcomes? In the example, the policy-maker might be responsible for the regulation of bus fares, and they may be able to encourage more off-peak travel with incentives for the bus operator.

The various ESD courses exposes us to different modelling tools. The modelling tools takes the context and the data and makes it abstract. For example, we can model the bus station as a queue, which a supermarket checkout counter can be modelled as too. This is how modelling techniques are measured

  • Performance. If we implement the suggestions we want to see measurable improvements.
  • Explainability. Given a phenomenon, we want to understand what For example, we observe that people report long wait time for buses. We want to understand the reasons behind the long wait for buses.
  • Simplicity. The model should be simple. Complex models may be able to explain more pheneomeons and have better performance, but the tradeoff has to be justify. Complex models also risk overfitting and may not be genealisable to greater concepts.

This is how I understand what my universities studies is about - on how it takes business problems and improve from it.