MSc Internships

Admission procedure for industrial projects.

If you are a MSc student in Computer Engineering (CE) interested in working in a project in the industry, please follow the procedure described below:

  1. Prepare a curriculum vitae (CV) and a course list with all grades (your GPA should be above 7).
  2. Send both CV and course/grade list to Zaid Al-Ars (
  3. Indicate the project(s) of your interest (see below). In case you find none, please contact Zaid Al-Ars for the most recent openings.

Your application will be forwarded to the company contacts and you will be notified if if your application has been accepted. An invitation for an interview can be part of the application procedure.

If you are accepted by the company:

  • Write a one page description of your specific project topic.
  • A CE supervisor will be assigned to guide you from the university side based on your topic.

It is strongly advised that the student has finished most course work before applying for an MSc project in the industry. Furthermore, international MSc students may require a 'work permit' before they can start their work. Obtaining such a permit may take approximately 2 months, so it is advisable to start looking early for an open position and to start your preparatory work while waiting for the permit to be issued. More information can be found on this page.

If you're interested in performing an internship in the industry and would like to get credit points based on your work, please follow the procedure as outlined on this page (see also course EE5010 on Brightspace). It is also possible for students to find industrial MSc projects by themselves. In this case, the project must be approved first. You should contact Zaid Al-Ars for more details on how to obtain approval.

MSc Internship Projects

[CE2018-5] AEGON (NL) - Neural Network Based Anomaly Detection for Insurance Fraud Identification

Benchmark studies estimate that around 5-15% of incoming insurance claims are fraudulent to some degree. As insurance companies handle hundreds of claims everyday, it is essential to process the correct claims fast to ensure customer satisfaction, and on the other hand double check suspicious claims extensively.  It has successfully been shown that anomaly detection can help intercept suspicious claims more efficiently. Literature shows that there are many different options for anomaly detection amongst which neural networks sound very promising.

The goal of this project is to design and build an unsupervised autoencoder neural network for anomaly detection. The model can be built in python or R.  In addition, a comparative study of neural network model interpretation methods should be implemented to help explain the predictions. The student will be working on real datasets from the field and gain experience in creating real data science applications for business.

Company Name: Aegon
Location: Den Haag, NL
Start date: Summer/Fall 2018
Duration: 3-9 months (Internship/MSc project)


[CE2018-4] TUDelft (NL) - Big Data Analytics for Evaluating Success Criteria of MSc Students in TUDelft

With the increasing number of students coming into the various masters programs in TUDelft, there is a need to improve the efficiency of the educational process and help students improve their chances of success in their studies. Students starting their studies at an MSc level face a number of new challenges, such as the freedom they have in building their own study program, selecting and managing an 8-month long MSc thesis project, as well as functioning at a higher professional level requiring technical writing, communication, and teamwork skills. In some cases, this causes costly delays for incoming students and might even result in students stopping their studies.

The objective of this project is use the available historical information on MSc students in TUDelft, and try to identify correlations between the knowledge of incoming MSc students and their success rates in their MSc studies. The work will take place in collaboration with the university student administration to create predictive models of these success criteria. This will enable the university of providing much better advice and guidance to incoming students to improve their chances of success.

Company Name: TUDelft
Location: Delft, NL
Start date: Spring/summer 2018
Duration: 1-3 months
Position type: Internship


[CE2018-3] Dutch Analytics (NL) - Big Data Analytics and Integration of Public Data Sources

Dutch Analytics specializes in the development of predictive maintenance solutions for major industrial players. We develop scalable software solutions that facilitate development, testing and implementation of machine learning models. Dutch Analytics kick-started in December 2016 and has since then been able to develop products and services, which have attracted interest from an increasing number of customers. An example of our applications is the prediction of railway switch failures in the Netherlands, which has one of the most dense railway networks in the world. Moreover, we are actively expanding our predictive maintenance applications into various other industries, both in and outside the Netherlands.

In this internship, you will research available open data for various applications, assess their value and build prototypes for integrating these data sets into our existing infrastructure. The number of publicly available data sources is increasing rapidly. Both companies and governmental organizations are opening up parts of their databases for everyone to make use of their stored datasets. This information is not proprietary but can be a very interesting addition to make some of our models more powerful. Examples are weather data, vehicle location data, schedules and benchmark data sets for asset performance.

Company Name: Dutch Analytics
Location: Delft, NL
Start date: Summer 2018
Duration: 2-3 months
Position type: Internship


[CE2018-2] Kempen (NL) - Modeling of machine learning applications for asset management

In recent years, the investment landscape has undergone tremendous changes, driven by a steep increase in data availability and computing power and significant developments in machine learning methods. Kempen, as a unique, specialist asset management company and a star player in its niche markets, is leveraging the opportunities to stay ahead in the international league in small caps and real estate, but also high-yield stocks, fixed income and funds of hedge funds.

The goal of this assignment is to create a machine learning model that contributes to alpha generation. As such supervised and unsupervised machine learning concepts varying from k-means clustering to reinforcement learning will be applied to enhance the bottom up investment process. The intern is responsible for the algorithm selection and programming. The assignment allows the intern to gain a thorough understanding of the investment dynamics.

The ideal candidate has a hands-on, pragmatic attitude, has significant experience with machine learning concepts and programming skills in Python, R and MatLab.

Company Name: Kempen Capital management
Location: Amsterdam, NL
Start date: Spring 2018
Duration: 3-9 months (Internship/MSc project)


[CE2018-1] Dot Robot (NL) - Software Development of Electric Mobility Scooter Control Systems

Dot Robot is a high-tech Yes!Delft startup located in the TUDelft. We have an engineering-driven team with great ambition. That means lots of work, but we also organize team activities. We have some great plans for the future, so we’d like students to push them forward, and we also like your enthusiasm to keep it fun and build a great team together.

We are looking for an intern engineer to work 1-2 months on the software development of an electric mobility scooter. The main task is implementing the drive controls and the state machine on the central control unit of the mobility scooter and testing the driving behavior. Optionally you will also implement a monitoring application, in which we can see live data from our test runs.

Company Name: Dot Robot
Location: Delft, NL
Start date: winter 2018
Duration: 1-2 months
Position type: Internship

[CE2017-5] Philips Healthcare (NL) - Modeling of Medical Visualization in Heterogeneous Node Architectures

Multiple medical scanners (CT, MRI, Ultrasound, x-ray, nuclear medicine) produce a stack of 2D images representing a 3D volume. To visualize such a volume, as a projection on a 2D image seen from an arbitrary angle, a so-called ray-casting algorithm is used. Our existing ray-casting algorithm is written in C and originally runs in a Window environment (Visual Studio 2013). This internship aims to develop a performance model to adapt the existing algorithm for further acceleration on a multi-processor machine with a variety of processing nodes. These processing nodes can be CPUs, GPUs and functions implemented in dedicated hardware (FPGAs).

The goal of this assignment is to create a model and determine its optimal parameters for distributing the kernel instances at the different node types that are available, aimed at reducing the latency. E.g. one of such parameters could be about the block size in relation to the cache sizes of different processing nodes. The existing code, although prepared for parallel processing, runs on a single machine and the proposed solution should be highly distributed. This means an effective and efficient distribution amongst the available heterogeneous nodes needs to be chosen. As the algorithm is memory bound, special care needs to be taken to distribute the input data amongst the nodes. A stack of images that make a 3D volume can be up to 1.5 GB. Other parameters are: bandwidth of the various interconnects, available FPGA area, size and availability of local cache memory.

Prerequisites: 1. Affinity with bottleneck analysis of a network of heterogeneous processing nodes and its interconnects, 2. Modeling capabilities of such a network, 3. Knowledge of Linux / gcc toolchain

Company Name: Philips Healthcare
Location: Eindhoven, NL
Start date: fall 2017
Duration: 3-6 months
Position type: Internship


[CE2017-4] Erasmus MC (NL) - Custom Numerical Precision of HPC Brain Simulations

The Neuroscience Department of the Erasmus Medical Center has a long experience with developing and accelerating high-performance and highly biologically accurate models of the brain (>1,000,000 neurons). To this end, they make use of various HPC technologies, including cutting-edge, FPGA-based Dataflow-Engines (DFEs) by Maxeler Technologies, as well as NVidia GPUs and Intel Xeons. However, due to the complexity and non-embarrassingly parallel nature of many brain models, designs have to be spread across multiple nodes of each HPC technology, which incurs a significant chip-to-chip penalty. Various solutions are being explored at the moment in Erasmus MC as well as in the larger HPC industry; yet, one crucial source of resource waste stems from the pure matching of numerical accuracy needed by the brain models and the one provided by the HPC platforms. The biologically accurate models handled in the lab do not allow for simplistic fixed-point arithmetic solutions but call for a careful accuracy and range analysis of our existing neuron-model dataflow graph. Once the multiple, different floating-point accuracy operators needed are identified, the existing model design must be customized with tailored math operators supporting different floating-point accuracies. Of course, attempting such an optimization on a software platform is impossible (e.g. GPU) or far too complicated (e.g. FPGA). We are thus making use of the DFE platforms which permit performing numerical exploration and automated synthesis of custom floating-point operators for an optimized brain-model dataflow graph. This thesis topic is a collaboration between Erasmus Medical Center and TU Delft (Computer Engineering, Applied Mathematics).

Company Name: Erasmus Medical Center
Location: Rotterdam, NL
Start date: summer 2017
Duration: 9 months
Position type: MSc project


[CE2017-3] Karolinska Institutet (SE) - Developing Image Processing Algorithm for Neurobiological Application

Karolinska Institutet is a leading medical university and research center. The Department of Neuroscience has a long tradition with various technical innovations, focusing mainly on the neurobiology of neuropeptides, with important contributions to the field of brain tissue and neurons. Currently, we have introduced the volume imaging technique iDISCO+ (3D immunostaining) that requires substantial computing expertise. Our laboratory has a light sheet microscope and accompanying computation platforms.

The intern will be responsible for optimizing and verifying the implementation of the volume imaging applications on available compute platforms. This will allow the intern to gain valuable insights into the development process in an industrial research lab. The ideal candidate will be a student working toward an MSc degree, with a strong background in computing and programming.

Company Name: Karolinska Institutet
Location: Stockholm, SE
Start date: Summer 2017
Duration: 3-4 months
Position: Internship


[CE2017-2] Xilinx Research (IE) - High-Level Programmability of Heterogeneous Multicore ARM-FPGA System

Xilinx is a world leading maker of FPGA devices. The internship will take place in Xilinx Research Labs Ireland in the Dublin site. The team's focus within the Research Labs in Dublin is to investigate the high-level programmability of signal processing algorithms and conduct experiments with new programming paradigms for Xilinx FPGAs, SoCs. In particular, current generation Xilinx Zynq devices with two ARM cores and programmable FPGA logic side by side are an interesting field of investigation.

The intern will be responsible for optimizing and verifying the implementation of the specified applications on these compute platforms. Additionally, he/she will gain valuable insights into the development process in an industrial research lab. The ideal candidate will be a student working toward a bachelor's or master's degree from an accredited academic institute, with a strong background in software programming. No detailed RTL/FPGA experience is required.

Company Name: Xilinx
Location: Dublin, IE
Start date: Begining 2017
Duration: 3-9 months
Position: Internship or MSc thesis student


[CE2017-1] Ximedes Software (NL) - Implementing Blockchain Solution for Secure Payment Transactions

Ximedes is a FinTech company, based in Haarlem, that offers software development services and transaction based solutions. As part of their effort to integrate new developments in payment transaction, they are interested in Blockchain technology as a promising technology for secure transactions. With the increasing popularity and interest in Blockchain, Ximedes wants to get hands-on experience by starting a proof-of-concept project for such implementation. The student is expected to already have knowledge of java, agile/scrum, micro-services and to be interested in Blockchain technology. Ximedes sees the following activities as part of the assignment:

  • Get a thorough understanding of Blockchain technology and the applications
  • Look into online projects as Ethereum en Blockstream
  • Look into Linux Hyperledger and Digital Asset Holding; as open-source platform for blockchain-applications. See also Multi-chain
  • Start with the development environment and the database
  • Implement a proof of concept application for closed payment platforms (, all based on permissioned ledgers.

Company Name: Ximedes Software
Location: Haarlem, NL
Start date: Begining 2017
Duration: 9 months
Position: MSc thesis student


[CE2016-3] Philips Healthcare (NL) - Tools and techniques for transforming software to portable hardware 

Philips Healthcare is constantly looking for methods to improve the quality, life time, cost price and maintainability of its systems. At our department we are working on systems which visually assist and guide physicians during an intervention. The X-Ray system should be reliable and should deliver real time high quality images to the physician.

Currently we are studying tools and techniques to transform a C/C++ based algorithm to configurable hardware while maintaining portability. You might, amongst other subjects, work on an implementation with OpenCL on a configurable multicore system on an FPGA. Both hardware and software knowledge are required.

This study is part of the European ARTEMIS ALMARVI project ( and part of the results will be presented in the study. You should have a background, but a least a strong interest and willingness to learn in:

- (Embedded) computer architecture
- Pipelined software and hardware
- Multicore systems and programming
- OpenCL
- C/C++ in general
- High Level Synthesis
- Microcontroller programming
- FPGA design
- Image processing
- The English language, both spoken and written

Company Name: Philips Healthcare
Location: Best, NL
Start date: End 2016
Duration: 3-9 months
Position: Internship or MSc thesis student


[CE2016-2] Technolution (NL) - Resource-friendly fault-tolerant memory management unit for the RISC-V processor

RISC-V is a new instruction set architecture that has promising features for both low-end as high-end applications. Technolution built its own RISC-V core targeting safety critical or security critical applications. These applications require the core to withstand or detect errors due to single even upsets (SEUs). In order to make the entire path from the core to the memory safe for these errors, we would like to have an MMU implementation that can detect (and possibly correct) SEU errors. Furthermore, we would like to add the MMU also for low-end applications. Therefore, the core needs to be scalable in the amount of resources used. Possibly only memory protection is performed for low-end applications.

The assignment is develop a resource friendly MPU & MMU implementation that is protected for SEU errors. The MPU/MMU is written in VHDL and targeted for FPGAs. The research questions of this assignment that need to be answered are as follows.

- What mechanisms are available to implement an MMU that has SEU protection in the data and control logic? Can both the control and data path use the same mechanisms for SEU protection or are different mechanisms more appropriate?

- Which architectural optimizations can be done to optimize the resource usage for the MMU/MPU when targeting low-end applications?

Company Name: Technolution
Location: Gouda, NL
Start date: End 2016
Duration: 9 months
Position type: MSc project


[CE2016-1] Erasmus MC (NL) - Multi-FPGA Implementation of Artificial-Cerebellum Computational Model

Over the last decade, an increasing amount of effort is being spent on constructing and, then, simulating powerful brain models that can greatly help in unraveling the mysteries of the human brain (see for instance the EU flagship project: Whereas these models are powerful and constantly come closer to the real brain functionality, however they are typically very computationally intensive, to the point that common platforms such as multicore CPUs fall short of reasonable execution times. We have thus turned to more powerful platforms, FPGAs. While FPGAs receive high marks when it comes to performance acceleration, nevertheless, their limited capacity is not sufficient for implementing large-scale brain simulations comprising (hundreds of) thousands neurons. The subject of this topic is to extend a currently implemented, biologically-accurate, simulation platform (comprising a single FPGA) to incorporate multiple FPGAs. If the inter-FPGA communication challenges are recognized and sufficiently dealt with, this extension is expected to double the achievable real-time brain simulation capabilities with every new FPGA on the stack. The platform is to be used for biophysically-meaningful simulations of Cerebellar microsections in the Neuroscience Department of the Erasmus MC, Rotterdam. The student is expected to analyze the original single-FPGA neural models, identify latency-sensitive sections and potential optimizations and, then, deploy (through use of suitable EDA tools, e.g. Compaan) the original application onto a multi-FPGA arrangement.

Company Name: Erasmus Medical Center
Location: Rotterdam, NL
Start date: 2016
Duration: 9 months
Position type: MSc project


[CE2015-2] Philips Healthcare (NL) - Multicore Implementation of Image Processing Algorithms 

Medical imaging devices are becoming more demanding, in terms of resolution, low latency, high throughput, etc. This sets ever increasing requirements on the computational systems used to process the information generated by these imaging devices. This project is concerned with designing and implementing cutting-edge image processing algorithms needed for high-end medical devices on heterogenous multicore platforms, such as CPU, FPGA and DSPs. The student will investigate the viability of these algorithms for multicore system and compare their performance between a number of system alternatives. Possible performance gains by running these algorithms on the rVEX processor will also be investigated.

This work is part of a collaboration between the TUDelft and Philips Healthcare within the ALMARVI European project. The results will be used to identify the best alternative platforms for next generation X-Ray imaging systems designed by Philips.

Company Name: Philips Healthcare
Location: Best, NL
Start date: Mid 2015
Duration: 3-9 months
Position: Internship or MSc thesis student


[CE2015-1] Philips Research (NL) - Acceleration and Optimization of Wide Area Ultrasound Communication Algorithms

New ultrasound communication/imaging techniques have enabled an increased throughput bandwidth for communication devices that require new efficient algorithms to ensure their appropriate processing. These algorithms need to be optimized and, in some cases, accelerated to meet their timing requirements. The student will combine knowledge about Digital Signal Processing (DSP), Matlab, as well as VHDL FPGA programming to create an appropriate ultrasound communication/imaging solution. The work is to be done within Philips Research in Eindhoven. Expected outcome includes the following:

  • Come up with suitable architectures for digital beam-forming for ultrasound in time and in the frequency domain
  • For the latter knowledge should be available for implementing FFTs, Hilbert transform (Cordic), FIR/IIR filters, ADC (SAR, sigma delta), CIC/SINC3 filters, I/Q demodulation, decimation/interpolation
  • Writing suitable test-vectors to perform verification on algorithmic level
  • Use the Matlab code to generate RTL-VHDL test benches and RTL-VHDL for the design blocks
  • Run synthesis on the RTL VHDL blocks to get an estimate on area/power/performance

Content and Goal: feasibility study of the concept and building a prototype based on existing ultrasound analog frontend + FPGA or DSP implementation

Student background: The student is expected to have knowledge of digital design, e.g. in application domains such as communications/imaging. Architectural design, having knowledge of MATLAB/VHDL programming and programmable DSPs or microprocessors (ARM, MSP). Having some knowledge on ADCs and analog circuit design or AMS could be a pre.

Company Name: Philips Research
Location: Eindhoven, NL
Start date: Mid 2015
Duration: 3-9 months
Position: Internship or MSc thesis student


[CE2014-4] Erasmus MC (NL) - Security in Body-Area Networks Using Heartbeat Monitoring

This topic is in the context of heartbeat-based security for implantable medical devices (IMDs). The time interval between heart beats contains a high degree of entropy, while it may be measured remarkably consistent throughout the human body. In other words, multiple entities on the same body may use this time interval to generate a nearly identical security key, showing minor disparities due to natural variations in the human body. Accordingly, secure communication is facilitated if the keys are "similar enough", evaluated by a key-classification scheme. In this topic, the student is expected to develop a key-classification scheme which adheres to the steep security requirements and tight constraints of IMDs.

Expected effort:
The student is expected to categorize various methods of heart-beat-based key-generation and key-classification schemes. Subsequently, these classification schemes are to be evaluated in terms of security and overheads (energy consumption, performance, etc). Based on this evaluation, the student is expected to design and evaluate a novel key-classification scheme tailored to heart-beat-based security.

Expected outcome:
A key-classification scheme which is suitable for heart-beat-based security in IMDs.

The student is expected to have a background in computer science, computer engineering or embedded systems and has a basic understanding of security concepts and low-power design.

Company Name: Erasmus Medical Center
Location: Rotterdam, NL
Start date: Start 2015
Duration: 9 months
Position type: MSc project


[CE2014-3] IBM (NL) - Detection of Security Vulnerabilities in Network Communication

IBM NL is a subsidiary of IBM based in the Netherlands, providing a diverse portfolio of computing solutions, services and products for large as well as small businesses in the fields of infrastructure, management, security, etc. This project is carried out in IBM Delft and focusses on research into the possibilities for improving the automated detection of security vulnerabilities in network communications setup of zSeries systems running z/OS, and sometimes z/VM with zLinux. Since these systems are often used in industries where information and its accuracy are of high value, attacks from legitimate users trying to exceed their authority as well as from outside can be expected.

Customers are worried about increased vulnerability as the knowledge and number of exploits available on the Internet increases. Concretely, under z/OS there exist a Policy Agent that can be used to define protection (like the type of encryption of authentication required) for network resources. There is however a need to verify the actual coverage of the policy over the network links that are really active. The network links are of two types: those implemented by (multiple) TCP/IP stacks, and those of SNA (Systems Network Architecture), with the added complexity of IP over SNA and SNA over IP being both possible. One of the troubles with analysing security is the limited understanding of the sensitivity of available resources. Some research into determining that to some extent in an automated way is warranted.

Company Name: IBM
Location: Delft, NL
Start date: End 2014
Duration: 9 months
Position type: MSc project


[CE2014-2] TOPIC (NL) - Facial recognition using Xilinx/Altera SoCs

In the domain of embedded processing, we see many different high-performance embedded processing platforms coming up. This varies from multicore processors via processors incorporating GPU and accelerator blocks to processors combining powerful CPUs with DSPs or FPGA fabric. Topic Embedded Systems is active in the area of embedded application development and very much focused on systems where software and FPGA functionality are overlapping. Think about video applications, algorithm implementation, data mining, etc. In this context, we want to explore the capabilities of the Altera and Xilinx SoCs, which both incorporate a dual-core Cortex A9 and a lot of FPGA fabric.

A typical application in our domain is facial recognition. The quality of facial recognition depends on the camera resolution, computational effort and applicable algorithm. Typically, the processor part of the SoC runs Linux. Part of the assignment is the implementation of a facial recognition algorithm that is able to recognize multiple faces in a video frame, track them over different frames and, when possible, identify the faces depending on a face matching database.

For the assignment you will use a USB connected camera as video source and an HDMI or a panel display for visualization. On top of Linux you will use e.g. Qt for visualization. The implementation of the facial recognition algorithm will make use of both the processor and the FPGA, exploring the capabilities the Xilinx or Altera SoC, the applicable tool flow and the available video processing functionality in terms of e.g. OpenCV, OpenCL, video tool kits, high-level synthesis (C to VHDL translation). The applicable SoC will be available as part of a Xilinx or Altera development platform. 

The result of the assignment will be a demonstrator, performing live facial recognition as well as a report on the investigated design routes, development flows and implementation experiences.

Company Name: TOPIC Embedded Systems
Location: Delft, The Netherlands
Start date: Start 2015
Duration: 9 months
Position: MSc student


[CE2014-1] MMM (NL) - Big Data Analyst

Big data applications are becoming ever more important to analyze the wealth of data being acquired from our environment. There is a lot of value to be gained by systematically organizing, classifying, and analyzing the collected unstructured data. For commercial organizations, this could translate to increased understanding of their respective market segments and identifying new product or market opportunities, which would translate to a financial benefit to those organizations.

You will work on investigating the required hardware and software infrastructure needed to build a big data analysis system. You will be working in a young and dynamic company to develop marketing solutions to big customers. The internship work will be carried out at MMM, Amsterdam (opposite to Amsterdam CS).

What do we expect from you?

  1. A data-driven, innovative, and hands-on attitude
  2. Analytical and critical thinking approach
  3. Experience with .NET
  4. Knowledge of data warehousing

Company Name: MMM (Make Marketing Magic)
Location: Amsterdam, The Netherlands
Start date: Mid 2014
Duration: 3 months
Position: internship student

CE Tweets