Master/Semester projects

Overview of available semester/master projects

The CHILI lab is inventing learning technologies that exploit recent advances in human-computer interaction (e.g. eye tracking, augmented reality, …) and in human-robot interaction. We are working on several educational platforms described below. Each platforms offers possibilities for semester and master projects. While semester projects are often limited to development, master projects usually include an empirical study with learners, supervised by our team.  The platforms are:

  1. NEW ! We have funding for supporting master theses in the field of learning technologies in Fall 2019 and Spring 2020. In 2017, EPFL has launched the Swiss EdTech Collider which now gathers 77 start-ups in this field.  Some of them will be interested to host master theses. You will be supervised by Prof. Dillenbourg or his team members but you will  located in a start-up (different cities in CH).  Contact: pierre.dillenbourg (at) 
  2. A variety of projects in LEARNING ANALYTICS, i.e.  data sciences applied to education are offered by my lab (contact: jennifer.olsen (at) as well  by the Center for Digital Education. See their project list here.
  3. REALTO and Training Needs Analysis are two parts of the project DUAL-T, which focuses on the Swiss Vocational Education and Training system. Realto is a social platform for vocational education. Apprentices collect pictures at the workplace and upload them on their class flow, where several picture annotation tools and augmented/virtual reality tools are available. Training needs analysis concerns itself with finding methods for identifying the newest skills needed for people in a profession, and involves the use of data science and applied machine learning. Current projects concern the augmented/virtual reality tools, as well as training needs analysis for software developers. Contact: catharine.oertel (at)
  4. CELLULO is a small robot for education. It moves by itself and can be moved by pupils. The hardware is ready and projects concern the software environments as well as designing and experimenting with new learning activities. Contact: hala.khodr (at), arzu.guneysu (at)
  5. CO-WRITER is a project in which kids who face writing difficulties are offered to teach Nao how to write. Nao is a small humanoid robot available on the market. The projects concerns smoothening the interaction between the robot and young children. Contact: thibault.asselborn (at)

Some of these projects are described below, but since research is moving on permanently, we always have new opportunities. You can always contact the names above or pierre.dillenbourg (at) if you are interested in advancing digital education.

Learning Analytics 

Learning analytics involves applying techniques in data science for optimizing and understanding learning. In CHILI, the projects range from applying existing algorithms to new data sets, comparing the use of algorithms to address a goal, and visualizing data in a meaningful manner to support learning. 

[Semester] Learning to explain the predictions of neural networks

The predictive powers of convolution neural networks in recent years have been achieved or surpassed human level performance. However, interpretability or lack thereof limits the adoption of these methods in decision critical domains. This project aims to address this issue in the context of diagnosing dysgraphia (a learning disability that results in an inability to write coherently). A popular solution is to apply saliency maps (eg. guided backprop and its variants). However, we think in terms of high level concepts, thus, pixel level activations can be hard to understand. We aim to learn an unsupervised representation of possible descriptors to be used as explanations.

Prerequisites: Deep learning, Pytorch, image processing

Contact: teresa.yeo [at], thibault.asselborn [at]

[Semester] Predict Student Learning Using Deep Knowledge Tracing

When students are working on a problem set, to be able to recommend the next problem that will best support their learning, it is important to understand their current state of learning. Within the field of education, this knowledge tracing, or understanding the probability that students have mastered certain skills, has primarily been done through Bayesian Knowledge Tracing. However, recent work has shown that deep learning may produce more accurate predictions.

In this project, you will work with a large data set of students working on algebra problems to apply deep learning methods to predict student knowledge and compare these results to other popular methods.

Prerequisites: Deep learning, Python or R

Contact: jennifer.olsen [at]

REALTO and Training Needs Analysis

REALTO is a social platform for vocational education. Apprentices collect pictures (and videos) at the workplace and upload them on their class flow. Supervisors and teachers have the possibility to provide feedback on the students private flow and peers have the possibility to comment on other students pictures and videos. Over 2000 apprentices from a wide variety of disciplines such as florists, carpenters, fashion designers are currently registered on REALTO.

Training Needs Analysis is the identification of skills that will help people in a profession improve their performance and obtain the skills they need. Currently, we are focusing on performing training needs analysis on software developers, for whom we have much publicly available data, in the form of Stack Overflow questions, Stack Overflow Developer Survey, and Google Trends.

Following are the list of available projects and their descriptions. In case of interest, please send an email to the contact person. In your email, please include your CV and a short description of your specific interests.

[Semester] The Stack Overflow Annual Survey: A look into the future? (Data Science)

Stack Overflow is a well-known Q&A website for developers, where users may ask questions and give answers about a wide variety of programming issues. Stack Overflow also runs an annual developer survey, where the participants answer questions about the programming languages and frameworks they use and the methods they have used to learn new languages. This annual survey happens in January and has been receiving over 60,000 responses each year since 2017, creating a rich dataset of what frameworks and languages developers used in the preceding year, what they will use in the next year, what kind of job they have, etc.

In this project, we look at data from the Developer Survey, along with data from Stack Overflow and Google Trends, to answer questions such as:

  • Are Google search volumes predictive of the changes that occur in programming language and framework usage, as indicated by the Developer Survey? In other words, how well can we predict the results of the next developer survey only using data that’s publicly available beforehand?

  • The question above, but inverted: are these transitions from one technology to another predictive of trends that will be seen throughout the year on Google Trends?
  • Do technologies that are becoming more popular also experience a surge in question counts on Stack Overflow? If that is the case for some technologies and not others, what are the common characteristics of each group of technologies?

Requirements are familiarity with Python, some experience in data analysis and applied machine learning, and preferably some familiarity with time series analysis. Having taken the course Applied Data Analysis (CS-401) or Lab in Data Science (EE-490) is a plus.

Contact: ramtin.yazdanian [at]

[Semester] A free MOOC market: What makes a course successful? (Data Science)

Udemy ( is a platform for sharing MOOCs, where skill-centric, self-paced MOOCs can be uploaded by anyone. It therefore essentially constitutes a free MOOC market, as opposed to most MOOC platforms which are controlled by specific institutions. This platform offers many opportunities for people who are looking to share their expertise with others, either for free, or for a fee.

In this semester project, you will use various data analysis tools (machine learning models, statistical tests, etc.) to investigate and understand the dynamics of Udemy. You will be looking to answer questions like the following:

  • Does the (relative) popularity of these courses mirror the (relative) popularity of their subject matter on Google Trends? In other words, is demand as represented by course subscription mirrored by demand as represented by search volumes?
  • What are the common traits of courses that end up being successful?
  • Are the earliest courses the most successful in their subject matter? What are the common traits of courses that get a head start but don’t end up being the most successful in their subject matter? 
  • What are common patterns observable in the appearance of courses on a new subject? Is there a process we can identify?

Requirements are familiarity with Python, general data analysis, and some applied machine learning. Having taken the course Applied Data Analysis (CS-401) or Lab in Data Science (EE-490) is a plus.

Contact: ramtin.yazdanian [at]

[Master] Deep design exploration

Description: On Realto, apprentices upload pictures taken from their workplaces (e.g., bouquets by florists or chairs by chair makers). We believe that these pictures can be a great source of design exploration. The idea is to allow the apprentices to explore design space by exploring variations of their own uploaded design. In this story, the goal of this project is to implement an algorithm that learns to generate design variations using convolutional neural networks. Our current interest is on a “chair” dataset.

Prerequisites: experience in following topics or interest in learning: deep learning, image processing, Python

Contact: (at)

[Master] Object detection using deep learning

Description: On REALTO, apprentices upload pictures taken from their workplaces and share them in the digital space. In order to make better use of the uploaded data, it is important to have some semantic understanding of the image. Recent advancement of deep learning algorithms has improved the performance on the problem of image-based object detection. The goal of this project is to implement, train and test state-of-the-art deep learning algorithms to recognize objects from images. We are currently working with a dataset of flower bouquets.

Prerequisites: experience in following topics or interest in learning: machine learning, image processing, Python

Contact: (at)


The CoWriter Project aims at exploring how a robot can help children with the acquisition of handwriting, with an original approach: the children are the teachers who help the robot to better write! This paradigm, known as learning by teaching, has several powerful effects: it boosts the children’ self-esteem (which is especially important for children with handwriting difficulties), it get them to practise hand-wrtiing without even noticing, and engage them into a particular interaction with the robot called the Protégé effect: because they unconsciously feel that they are somehow responsible if the robot does not succeed in improving its writing skills, they commit to the interaction, and make particular efforts to figure out what is difficult for the robot, thus developing their metacognitive skills and reflecting on their own errors.

[Semester] CoWriter setup with the QtRobot

In the CoWriter project, the children help the robot to write better and in doing so, they improve their own handwriting. Here, we use the “learning by teaching” paradigm which is motivated by the fact that being in the responsible role of ‘teaching’ boosts the self-esteem of the children who already can have confidence issues if they face handwriting problems in the classrooms. Until now, for the robot in the current setup, we have been using a NAO. The goal of this semester project is to make use of a QTrobot instead of NAO. Specifically, the task is to transfer the CoWriter project from Nao to QTrobot and to utilize emotional expressive behaviors, verbal and non-verbal, to probe and maintain engagement among the children interacting with the setup. The prerequisites for applying include experience in the following skills or interest in learning: ROS, python, git, etc. A small user study may be conducted at the end to validate the CoWriter setup with the QTrobot.

Contact: thibault.asselborn (at) or jauwairia.nasir (at) or utku.norman (at)


In the Cellulo Project, we are aiming to design and build the pencils of the future’s classroom, in the form of robots. We imagine these as swarm robots, each of them very simple and affordable, that reside on large paper sheets that contain the learning activities. Our vision is that these be ubiquitous, namely a natural part of the classroom ecosystem, as to shift the focus from the robot to the activity. With Cellulo you can actually grab and move a planet to see what happens to its orbit, or vibrate a molecule with your hands to see how it behaves. Cellulo makes tangible what is intangible in learning.

[Master/ Bachelor Semester] Cellulo: Pacman Game with Dynamic Workspace

In the Cellulo project, we are designing tangible robots to be used in games. Our robots operate on tabletop paper sheets and are used as game elements where they can be physical input devices and/or autonomous agents. Our hypothesis is that we can build tabletop games with these robots that can move and be moved, possibly at the same time. One of our current goals is to explore game design options that create engaging interactions between multiple players through these tangible robots and shareable game spaces. Our robots work on printed paper sheets that can be produced with up to ~1m width and with unlimited length, and so far we have used these large shared workspaces within our activities to promote scalable multi-user interaction. We have recently developed smaller workspaces that can be tiled dynamically during runtime, allowing the growing, shrinking and modification of the workspace and its shape by the players.

The goal of this project is to design a tangible game that uses this element at its core, together with the Cellulo robots. The resulting game will require the players to build the shape and the functionality of the workspace itself as the game progresses, on which their robots will move and be moved. A small user study will be conducted at the end to validate the player interaction with the developed game.

  • Extend Pacman game to two players or more
  • Extend Pacman with tiles to have a dynamic game space

Prerequisites: Experience in the following skills or interest in learning: Qt/QtQuick development, QML programming, git, game design,
Contact: arzu.guneysu (at)

[Master/ Bachelor Semester] Cellulo: Human Arm Activity Estimation During Pacman Game-Play

In the Cellulo project, we are designing gamified rehabilitation activities with tangible robots. The unique functionalities of Cellulo (i.e. haptic feedback and submillimeter precision localization) make it an interesting device for upper-arm rehabilitation. One of the designed games is tabletop Pacman game where the user performs upper limb activities while playing against ghost robots. This Pacman game has been tested with several healthy and impaired participants. Among them, the game-play data of 33 healthy participants and 10 impaired participants were recorded with a 2d and 3d camera. This semester project will investigate the construction of the skeleton data of the players during game-play and calculation of angles corresponding to shoulder, elbow and wrist motions of the player through video data. The final goal will be finding a correlation between end-effector (Pacman robot held by the user) position and the arm angles. Therefore, the game system will be able to make the estimations of performed arm activities through Pacman data.

The project will consist of:

  • Extraction of upper skeleton data through 2d or 3d video data (by using open pose library: ) and calculation arm angles.
  • Investigating the relationship between Pacman pose data (end-effector position) and corresponding arm activities such as angle changes in elbow and shoulder
  • Estimating the performed angle values of joints through Pacman pose data.

Prerequisites: Experience in the following skills or interest in learning Computer Vision, Python, C++, ROS, QtQuick, Data Analysis, git.
Contact: arzu.guneysu (at) , hala.khodr (at)

[Master/ Bachelor Semester] Charging station for Cellulo

In the Cellulo project, we are designing tangible swarm robots to be used in elementary school classrooms for learning activities. The goal of this project will be to create a charging station for the Cellulo robots to ease the mass charging and on-off control of a Cellulo Swarm.
The desired outcome would be a charger base connected to a power supply where a group of Cellulo robots can be placed to be charged and a switch to trigger the robots ON.

The student will :

  •  study the different alternatives for the most suitable contact design.
  • modify the power circuitry accordingly.
  • build a prototype

Prerequisites: Experience in the following skills or interest in learning: electronics design, mechanical design, prototyping

Contact: hala.khodr (at)

[Master/ Bachelor Semester] Modular Cellulo top:

In the Cellulo project, we are designing tangible swarm robots to be used in elementary school classrooms for learning activities. In its current version, the Cellulo top has illuminated touch buttons. In this project, we would like to have a modular top which will allow us to add/change peripherals, such as an illuminated touch buttons, a force sensor, a 9-axis IMU, a gripper, an ACM. The student will:

  • Re-design the top to allow an easy modular change
  • Design the electronics and printed circuit board necessary for the communication to the main microcontroller.
  • Choose one example of a peripheral and Implement the firmware to demonstrate the idea

Prerequisites: Experience in the following skills or interest in learning. Embedded systems design, electronics design, mechanical design, prototyping

Contact: hala.khodr (at) , arzu.guneysu (at)