Automated transaction classification using machine learning algorithm

The procurement process of an organization is key to understand company costs. Organizations gather large amounts of data coming from different sources (e.g. income statement, balance sheet, general ledger lines). This information is heterogeneous in nature as it is a mix of unstructured and structured data. Moreover, it needs to be cleaned and consolidated in a taxonomy to enable category management. The objective is to group like-to-like items and/or services into categories from Supply Market Analysis point of view and consider category management for the holistic spend.

Exploring optimal trading rules in a high-frequency portfolio

Given a set of financial instruments with inherent characteristics at different time intervals, we are interested in finding an optimal trading rule in a high-frequency trading context. A trading rule is defined as a combination of indicators as well as an entry threshold (and potentially other trading parameters). The objective function we are trying to maximize is the profits of the strategy based on the trading rule. One impact of the non-linearity of such problems is that the gradient of the objective function is hard to estimate using a black-box approach.

Preclinical proof of concept study evaluating PBI-4050 and analogs as potential treatments in the context of ischemic cardiomyopathy

The ability to impede/reduce complication of the damaged heart presents a major challenge in the treatment of cardiovascular diseases. Complications include heart failure, which has a high mortality even with current treatments. The use of a new drug to stimulate protection of the heart during an ongoing myocardial infarct and long term changes leading to heart failure would be very relevant to the clinical setting, to help patients suffering from diverse heart problems.

Longitudinal Weak Labeling for Lung Cancer Prognosis and Treatment Response Prediction

This project aims at evaluating whether recent results in deep learning models, trained to exploit weak labels (Hwang, 2016) can serve to extract meaningful lesion localizations from image-level labels, either from individual scans or given a (longitudinal) sequence thereof. To this end, we will scale up existing models that have been shown to work on 2D images to a 3D context, studying labeling performance as the dataset size grows.

Characterization of molecular pathways mediating the effects of novel therapeutic agents in heart failure

Heart failure (HF) is a condition that develops after the heart becomes damaged or weakened. HF occurs when the pumping action of the heart is not strong enough to move blood around, especially during increased activity or under stress. In addition, the heart muscle may not relax properly to accommodate the flow of blood back from the lungs to the heart. These abnormalities in heart function can cause fluid to back up in lungs and in other parts of body.

Efficacy of a novel anti-IL-1B receptor modulator in reducing preterm birth impact on neurovascular health

Preterm neonates ill-adapted to the extra uterine environment are prone to increased inflammation in multiple organs and the proinflammatory interleukine IL-1b has been closely implicated in brain injury associated with preterm birth (PTB). One major adverse neuronal outcome for PTB survivors is the greater propensity to develop ischemic brain lesions long after birth. Here, we hypothesize that the neural vasculature of premature infants becomes maladapted to appropriately respond to hypoxic-ischemic stress.

Implementing Factor Models in Investment Management

The internship will consist of studying, building, implementing and testing so called factors that are used to characterize the equities, commodities and currencies that the company invests in. These factors can be thought of as characteristics relating a group of securities that is important in explaining their returns and risk. My task will be first to understand the risk factors that are of particular importance to the company’s investment strategy.

Raman micro-spectroscopy for biopsy for prostate cancer prognosis -Year two

Prostate biopsies can be difficult to interpret using standard pathology techniques. Because cancer can be fatal, the development of technologies providing complementary information could improve current pathology practice resulting in improved patient outcome. Raman micro-spectroscopy is a molecular imaging technique using backscattered light following tissue laser excitation to indicate whether or not prostate samples contain cancer cells. However, this technique is very sensitive to the molecules in chemicals used to process tissue (e.g. paraffin), which greatly limits its efficacy.

Characterization of Naturally-Occurring Neuropathic Pain in Dogs

Clinical experience demonstrates that canine patients commonly suffer from neuropathic pain and little is known to address this issue. Our study aims to investigate different tools for the diagnosis and treatment of neuropathic pain. Forty dogs with naturally-occurring neuropathic pain will be included in a prospective, randomized, masked clinical trial using appropriate inclusion and exclusion criteria.

Mycobiota dysbiosis in colitis-associated colorectal cancer

A higher incidence of colon cancer is reported in populations consuming high amounts of red meat, as well as in patients with inflammatory bowel disease, where gut bacteria participate in the development of inflammation. We found that dietary supplementation with heme, an element found at high levels in blood, is detrimental to gut health and it fosters the growth of harmful bacteria. In this project, we propose to establish new procedures for both the manipulation of the gut bacteria (microbiota), as well as its characterization using a new state of the art sequencer.