13 Oct 2020
Oxford placements go virtual for three students from Centrale Nantes engineering school
Three French students joined the Department virtually this year to work on projects relating to in-networking computing and AI for healthcare, working with leading academics
Three final year engineering students from the Centrale Nantes school in France were given the opportunity to work with leading Oxford academics on research projects from April to September this year.
The industry-funded placement scheme has been running since 2019, with students from participating French engineering schools, or “Grandes Écoles d'Ingénieurs”, applying for carefully selected research projects within the Department. As in 2019, this year’s placements were intended to take place in Oxford, however due to the restrictions imposed by the Covid-19 pandemic, they were quickly reformatted to run virtually. Given the circumstances, placement facilitator Professor Antoine Jérusalem was pleased the internships were still able to go ahead: “Unfortunately, it was not possible to have the students coming here, but I think we can still consider the fact that we had three interns working remotely a success, in view of the situation.”
Alexandre Guerin, sponsored by EDF Energy, worked with Professor Jens Rittscher and Dr Tapabrata Chakraborty to develop AI models that aim to diagnose tumorous areas on histology slides, specifically in relation to colorectal cancer (CRC). Colorectal cancer is the 4th most common cancer in the UK, with survival rates strongly influenced by the speed and timing of diagnosis. Conventional screening methods are often time-consuming and expensive. The models built by Alexandre to describe the texture of histology slides showed encouraging results, as well as those designed to spot cancerous areas.
He says, “To fulfil my research, I have been granted the opportunity to use the BMRC cluster [Biomedical Research Computing cluster, partnered with the Big Data Institute]. It is a great help, especially when it comes to handling large biomedical datasets and to perform the training of my models, yet also a responsibility as many people rely on it to conduct their own work.” Having specialised in applied maths and computers during his general engineering course and with previous experience in software engineering and data analysis, he found the research experience very beneficial and it is helping to inform his decision whether to do a PhD, by showing how theoretical concepts are applicable to real life situations.
Damien Alliot, sponsored by an individual donor, worked with Professor Philip Torr and others in the Torr Vision Group in machine learning. He says “My work consists in finding a good way to read letters in an image and extract a representative structure of each character, that is pen strokes that could lead to the original image if a human were to draw it.”
Damien found ways to work around the limitations of working remotely and has developed new work skills as a result: “The main problem is the communication, and it required a bit of adaptation. I needed to take the habit of writing things down in Slack every day. It was a bit annoying at first, but then you quickly realize that with your supervisor being able to react to your progress and giving you appropriate directions to follow, you end up saving a lot of time. Besides, putting things into words is a good exercise to make sure everything is understood. These things were not difficult to set up, it was just a matter of good habits that I didn't have because it all happens naturally when you can discuss freely in the workplace and not so much in remote working conditions”.
Antoine Bernabeu (also sponsored by EDF Energy) worked in the emerging research area of in-network computing, i.e. computing within network devices, and explored the applicability to language processing applications (such as on-the-fly spell-checks and translation). He was supervised by Professor Noa Zilberman.
Offloading applications typically run on servers to programmable network devices, previously an unexplored area, can bring incredible performance benefits in terms of throughput increase (x10,000) and latency, and is a thousand times more energy efficient than using CPUs. However, the architecture of network devices was not designed to support such applications.
Using specialist programming languages Antoine completed the design of algorithms for spell-checking of text on-the-fly and tested them on the 2 billion words in Wikipedia. This required him to study new networking and hardware materials and enabled him to gain experience using cutting-edge technology. The new methods were shown to run successfully in terms of functionality, resource use and performance – with potential savings of 80% of host processing.
He says, “Under the supervision of Professor Zilberman, I have developed an application that performs spell checking while data is transferred within the network, using a lexical-dictionary. This is the first time that this kind of an application is running within the network, and it paves the way to larger projects, combining more complex solutions. In-network computing is opening a door to a whole new playground for programmers. Our work is a step in the direction of a more and more programmable network.” He concluded that the project was successful in demonstrating a proof of concept for further research in this area.
Despite not being able to come to Oxford as planned and having to carry out their placements virtually from France, the students found it overall a rewarding experience and it has enabled them to complete their final year studies. Antoine reflects, “I was not a specialist of networking and with this internship, I learnt a lot in this area. Regarding remotely, in my opinion it is not the best way, however I think it went ok. I would like to thank Professor Zilberman for her support during this placement.”
Alexandre also found the Covid-19 restrictions had an impact on his placement, saying “It's hard to make plans for the near future. My wish since the start of the project has always been to work on-site, because nothing replaces the interactions you can have in a normal day with your colleagues. Exclusively working remotely since April made a lot of things harder: communication with my tutors required for instance to find common time slots in busy schedules, so moments where I could ask for help or validation have been scarcer than expected.” However he adds, “I'm glad of the work that has been done and of the efforts I've never ceased to put into it”. He feels the placement gave him a lot of experience to take forward to a PhD in computing.
The Department intends to expand the placement scheme in 2021, with up to 4 students joining the Department from other Grandes Écoles d'Ingénieurs (Engineering schools) in France. Professor Jérusalem adds, “As well as giving the students valuable further insights into state-of-the-art engineering research in Oxford, it offers a unique platform to explore new research avenues, potentially leading to a full-scale PhD or similar projects”.