Paris/London - Long-term Internship - Global Markets - AI & Data Lab
Global Market is currently recruiting talented people to join one of the most challenging and exciting part of our Quantitative Research team, the Data and Artificial Intelligence Lab !
We are currently recruiting interns (Paris and London) for the Global Market Data and Artificial Intelligence Lab of BNP Paribas: Global Market is part of the Corporate and Investment Bank and deals with all market activities on Equity, Foreign Exchange and Local Markets, G10 Rates, Primary and Secondary Credit and Financing asset classes.
The Lab mission is to leverage the latest techniques of Machine Learning (Deep Learning, Natural Language Processing) on the vast amount of structured and un-structured data we are collecting while doing our business as well as any other public source of information.
We are, among other things, building models to improve the service we give to our Clients (issuing recommendation, anticipating their needs, bringing the relevant research…), to help traders better understanding and managing their risks or leverage alternative data sources (social media, news, images…) for the benefit of our strategists.
We are looking for candidates with education in data science, who not only have experience in solving complex problems but as well understand how and why the model work the way they do.
They need to be motivated with dealing with large amount of very diverse data and extracting valuable insights out of it.
The right candidate needs to be able to adapt quickly to new challenges, not to be afraid to experiment many times and fail before finding the right solution, challenge themselves with the feedback of the users and they will have the excitement of seeing their work being used in real live by the business.
For internships, we are looking at duration between 3 and 6 months and we are flexible on the starting date (the earlier the better!). The intern will participate to the life of the LAB and will take ownership of one or more topic. We have a great variety of topics, and you can find below some initial propositions:
- Prediction of which products are the most likely to be interesting for a given Client: through Deep learning network, we will combine techniques of time-series modeling and of collaborative filtering to do our prediction model. We will have to leverage not only structured data but as well un structured data like news or other text like information on our client interests
- Automated Generation of Market Comment: using Deep Neural Network in the context of Natural Language Generation, we will teach a network to write market comments based on what has happened during the day (Market Moves, transactions done with our clients, News…).
- Analysis of sustainability of corporates: delivering on sustainability is a license do to business in the 21st century and BNP Paribas is proud of its commitments in its space. The goal of the project will be to use all kind of data (news, social medias, and fundamentals on corporate, rating agencies…) to assess the situation of a corporate from a sustainable angle. In particular we want to be better at anticipating a potential reputational crisis to adjust the strategies we offer to our clients. The project will leverage Deep Neural Network to analyse market sentiment and anticipate future market movements. We may as well develop Chat Bots to interact with Sales or Clients on this topic.
- Optimal Risk Management of Interest Rates Swap Risk: we will teach a network on how to best manage a position of Interest Rates Swap by covering the risk of some less active instruments with more active ones. We will use neural networks calibrated potentially through reinforcement learning techniques
- Computer Vision: one of the key challenges we have is to be able to deal with the vast amount of documentation we process on our products or clients: one of the key source are the PDF documents which are not easy to deal with: we would like to create a network who could extract tables, comments, paragraph or sections in a very structured way. This will be done through latest computer vision neural networks.
- Word2Vec extension to supervised tasks: Instead of calibrating w2v on the full corpus, we would train the embedding of the words on our specific NLP task and at the same time (multi-task learning) calibrate the embedding to minimize the same loss as word2vec. This will allow the model to represent words that are in the training set of the supervised task correctly and give embedding that are in the larger corpus to be “close” to the one learned in sample.
- Quantitative Investment Strategies: one of our successful business is to propose investment strategies to our clients, a bit like hedge funds would do, but in a more transparent manner: indeed all the investment strategies are to be based on algorithm and the rules of the algorithm have to be described explicitly to the client. We would like to offer a new range of product based on Artificial Intelligence where our investment strategies benefit from the extra predictive power of Neural Network but as well benefit from other sources of data like News, market sentiment, fundamental data on companies… Models will leverage LSTM or CNN architecture to understand time structure of the data, and as well to process the unstructured sources of data.
- More computer science project: we would like to have a better understanding on where GPU outperforms CPU: the goal of the internship will be first to benchmark the two on various problems (regression, classification in the context of prediction or Natural Language Processing) and then to optimize the networks to fully benefit of the GPU power. A second part of the internship could be then about developing a Machine Learning framework where for a given model we automatically test various type of models, and various standard data processing pipelines in order to optimize a given score. This optimization will be done given a certain power/time constrain and will leverage various optimization techniques (random search, genetic algorithms or bayesian approaches)
Please send your applications to GM EARLY CAREERS firstname.lastname@example.org, indicating desired start date and length, preferred location and whether or not you have a strong preference among the potential topics.
This programme is closed to applications.