Manager, Data Science
Thomson Reuters - Other Locations
Total years of experience :8 years, 10 Months
Key Responsibilities:
• Contributing to technology planning & direction, strategy development, implementation, and delivery of AI solutions for real-world business problems in legal, tax & accounting, corporate and compliance domains
• Coordinating with product managers, stakeholders, and clients to ensure satisfactory fulfilment of technical and business requirements
• Identifying novel relevant technologies, trends and best practices and promoting the usage of the same across the organization
• Spearheading the creation of cutting-edge data science capabilities that enhance the business’s long-term performance by ensuring technological competitiveness
• Leading, mentoring and line management of a team of data scientists working across different projects
• Actively contributing to the rapid growth of a global data science team (20 new hires) by designing and conducting technical and behavioural interviews and identifying top talent
Significant Accomplishments:
• Developed and delivered seq2seq deep learning solutions for automated information extraction from millions of legal and tax documents (five models in production, cutting millions of dollars in operational costs for the business)
• Delivered a transformer-based AI feature for information retrieval and search, resulting in 300% increase in product sales
• Led the modernization effort for information retrieval and entity matching using big data in one of the company’s major products
• Led the development of unsupervised algorithms for the identification of financial and billing anomalies in millions of client invoices (one model in production)
• Leader for the “NLP for professional documents” research track within R&D organization
• Designed and conducted double-blind surveys for human evaluation of machine learning and deep learning solutions
Significant Accomplishments:
• Developed an agent-critic reinforcement learning framework for de novo anticancer compound design via a combination of a generative model (a Conditional Variational Autoencoder) as the agent and a multimodal attention-based predictive model as the critic (resulted in an invention disclosure and two publications)
• Designed an application with word2vec/skip-gram models in combination with high-dimensional clustering of word embedding vectors for the inference of entity interaction networks from the corpus of PubMed abstracts
• Spearheaded development of predictive machine learning models (e.g. Convolutional and Recurrent Auto Encoders) for unlabelled time series data in applications related to wearables and IoT systems for occupational stress detection (resulted in a publication)
• Led the development of multimodal attention-based deep neural net models for drug discovery in cancer using a plethora of multimodal data e.g. gene expression data, drug chemical fingerprints, SMILES string encodings as well as available prior knowledge protein-protein interaction networks (PaccMann project: https://www.zurich.ibm.com/paccmann, resulted in 3 publications & invention disclosure)
• Devised a Network-based Biased Tree Ensembles (NetBiTE) algorithm that exploits multiple sources of prior knowledge and network propagation in order to make faster and more accurate predictions of drug sensitivity while identifying crucial predictors for drug sensitivity in cancer (resulted in a publication and an invention disclosure)
courses: - Received the top-tier NSERC Alexander Graham Bell CGS-D award
Consistently received the annual departmental top-GPA recognition award - Conducted research on numerical simulation and control of a hydraulic robotic arm (
Consistently received the annual departmental top-GPA recognition award - Conducted research on numerical simulation and control of a hydraulic robotic arm (
SELECT PATENTS AND PUBLICATIONS 1. Skylaki L., Oskooei A., Gweder M., Bari O., Herger N., Kriegman Z.,