Abstract:
This talk will discuss the Exposing.ai research project, a multi-year investigation into the origins and endpoints of biometric image training datasets created from “media in the wild”. Over the past years several prominent datasets, including MS-Celeb-1M, DukeMTMC, VGGFace2, and MegaFace have been retracted, heavily criticized, or mysteriously deprecated without explanation. Driving these takedowns is the research project exposing.ai that has uncovered how and why academic biometric datasets are being exploited by the global biometric surveillance industry. As a recent [Nature article](https://www.nature.com/articles/d41586-020-03187-3) has explained, many researchers are reconsidering their use of these datasets and whether it’s appropriate for academia to engage in this research. This talk will discuss several of the most egregious dataset missteps and provide a survey of recent trends and papers about improving dataset authorship.

What You’ll Learn:
How to be a better dataset author

Adam Harvey (US/DE) is a researcher and artist based in Berlin focused on computer vision, privacy, and surveillance technologies. He received his masters degree from the Interactive Telecommunications Program at New York University (2010) and a BA in Integrative Arts from Pennsylvania State University (2004). His previous work includes CV Dazzle (camouflage from face recognition), the Anti-Drone Burqa (camouflage from thermal cameras), SkyLift (geolocation spoofing device), and Exposing.ai (interrogating face recognition datasets). His art and research has been featured widely in media publications including the Economist, New York Times, Financial Times, Süddeutsche Zeitung, Der Spiegel, Wall Street Journal, and the Washington Post. Harvey is the founder of VFRAME.io, a software project that innovates computer vision technology for human rights researchers and investigative journalist, which received an award of distinction from Ars Electronica and nomination for the Beazley Design of the Year award in 2019.

Abstract:
One of the top challenges in AI/ML is the black box models cannot be trusted in high-risk areas due to lack of explainability. Generally speaking, Explainability in ML is two folded, Casual Explainability (also known as interpretability) and Counterfactual Explainability. While the former addresses ‘why’, the latter addresses ‘how’ small and plausible perturbations of the input modify the output? The author’s focus is on Counterfactual Explainability from the optimization lens.
In ML, the learning phase is actually a constrained optimization problem where a given objective (a.k.a, loose) function must be optimized in terms of some constraints, e.g., regularization, lasso, dropout, etc. Thus, from the constrained optimization lens, the explainability in fact refers to ‘sensitivity analysis’ or ‘post optimality’ practice.
Using post optimality, we should focus on those learning coefficients that have a narrow range of optimality and coefficients near the endpoints of the range. However, the key conjecture in post optimality is: the optimization (learning) algorithm guarantees the ‘global’ or near-global optimum. Indeed, the majority of optimization algorithms in ML cannot guarantee the global optimum due to uneven (non-convex) loos surfaces or stochastic nature of the method. Actually, non-convexity and stochasticity are two sides of the complexity coin.
In this talk, the author argues that there is a trade-off between model explainability and accuracy. The lack of global optimum guarantee is the key reason why high-accurate (and mostly black box) models are not explainable. The above trade-off raises a critical discussion during the mode selection phase: a more explainable but less accurate model is better than a less explainable but more accurate model!?

What You’ll Learn:
There is a trade-off between model explainability and accuracy

Nima has a Ph.D. in system and industrial engineering with a background in Applied Mathematics. He held a postdoctoral position at C-MORE Lab (Center for Maintenance Optimization & Reliability Engineering), University of Toronto, Canada, working on machine learning and Operations Research (ML/OR) projects in collaboration with various industry and service sectors. He was with Department of Maintenance Support and Planning, Bombardier Aerospace with a focus on ML/OR methods for reliability/survival analysis, maintenance, and airline operations optimization. Nima is currently with Data Science & Analytics (DSA) lab, Scotiabank, Toronto, Canada, as senior data scientist. He has more than 40 peer-reviewed articles and book chapters published in top-tier journals as well as one published patent. He also invited to present his findings in some ML top conferences such as GRAPH+AI 2020, NVIDIA GTC 2020/2021, and ICML 2021.

Abstract:
In the past half-century, advances in computation have accelerated & scientific progress and innovation in diverse fields at all scales, from particle physics to socioeconomics to cosmology. However, recent works point to stagnating innovation and diminishing returns in science (Cowen & Southwood ’19). Can advancements in artificial intelligence (AI) and machine learning (ML) help push science, engineering, and other fields through existing bottlenecks (be they physical, computational, or otherwise)? Is AI-driven science necessary for humankind to decode and solve its greatest challenges, such as nuclear fusion and neurodegenerative disease? The application of AI in scientific discovery presents very different challenges relative to popular environments such as game-playing and machine translation. In general, scientific discoveries require hypothesis and solution spaces that are orders of magnitude larger than existing AI environments, a far more elaborate verification process, and non-trivial integration with scientific materials and machines. In this talk, leading AI researcher Alexander Lavin will discuss the challenges and opportunities in AI-driven science, and further propose a “Nobel-Turing Challenge”: AI systems capable of making Nobel-caliber discoveries in science. Lavin will present several key areas of AI/ML, simulation, and computing to advance towards this goal, recent progress in areas such as chemistry and climate, and critical operational aspects like human-machine teaming and systems engineering in AI.

What You’ll Learn:
The aim is to share the opportunities and challenges at the intersection of AI and science, particularly in very timely and relevant sectors such as energy, biology, and climate.
Researchers and engineers in AI/ML and scientific fields will take away an understanding of key technologies for advancing scientific AI/ML, as well specific methods or concepts to apply in their own work. Leaders and entrepreneurs will take away inspiration and practical directions to align their initiatives (companies, teams, etc.) with the pursuit of AI/ML for science. Overall, the presentation of a “Nobel-Turing Challenge” will intrigue and motivate audiences broadly.

Alexander Lavin is a world-leading AI researcher and software engineer, specializing in probabilistic machine learning, scientific computing, and human-centric AI systems. He is the Founder and Chief Technologist of the Institute for Simulation Intelligence, a public-benefit “focused research organization” aiming to reshape the scientific method for the machine age, building novel technologies to synergise AI and simulation in areas such as climate, synbio, nuclear energy, and more. Lavin has explored AI and probabilistic computation via several perspectives: theoretical neuroscience and online learning with Numenta, general intelligence in robotics and computer vision with Vicarious AI, predictive medicines and causality as the founder of Latent Sciences (acquired), Earth systems and climate with NASA as an AI Advisor, and autonomous systems with Astrobotic. Lavin earned his Masters in Mechanical Engineering at Carnegie Mellon, a Masters in Engineering Management with Duke University, and Bachelors in Mech & Aero Engineering at Cornell University. He has won several awards for work in rocket science and space robotics, published in top journals and conferences across AI/ML and neuroscience, and was an honoree for the Forbes 30 Under 30 List in Science and the Patrick J. McGovern Tech for Humanity Prize. In his free time, Lavin enjoys running, yoga, live music, and reading sci-fi and theoretical physics books.

Abstract:
Artificial Intelligence (AI) is changing how we live and how we work. As we advance into the 4th Industrial Revolution, the digital age, AI enabled technologies have the power to be a force for good. These technologies can also amplify forms of inequality, discrimination, and bias. This presentation will highlight cases of how algorithmic bias happens which can have significant impacts to society.

What You’ll Learn:
We will discuss how each of us can participate in influencing fair, equitable and inclusive technology.

Abstract:
This talk will feature excerpts from my recently published book “Human-in-the-Loop Machine Learning: Active learning and annotation for human-centered AI”. I’ll cover some of the most exciting problems in Human-in-the-Loop Machine Learning and promising recent advances that address some of these problems. The talk will start with one of the most basic and long-standing questions in machine learning: what are the different ways that we can interpret uncertainty in our models? The talk will then discuss recent advances in transfer learning, including active transfer learning for adaptive sampling and the implications of intermediate task transfer learning on the choice of annotation task and annotation workforce(s). Finally, I will talk about advances in annotation quality control and annotation interfaces, including ways to identify annotators with rare but valid subjective interpretations and human-computer interaction strategies for combining machine learning predictions with human annotations.

What You’ll Learn:
You will learn about how to create good training data for machine learning while also learning about some of the interesting open problems in the fields of active learning and annotation.

Robert Monarch is an expert in combining Human and Machine Intelligence, working with Machine Learning approaches to Text, Speech, Image and Video Processing. Robert has founded several AI companies, building some of the top teams in Artificial Intelligence. He has worked in many diverse environments, from Sierra Leone, Haiti and the Amazon, to London, Sydney and Silicon Valley, in organizations ranging from startups to the United Nations. He has shipped Machine Learning Products at startups and at/with Amazon, Apple, Google, IBM & Microsoft.
Robert has published more than 50 papers on Artificial Intelligence and is a regular speaker about technology in an increasingly connected world. He has a PhD from Stanford University. Robert is the author of Human-in-the-Loop Machine Learning (Manning Publications, 2021)

Abstract:
With advancements in technology, specifically around an organization’s ability to make sense of data, consumers are growing more accustomed to personalized, on-demand solutions. In business like HelloFresh where the objective is to offer just that – solutions – product design and re-design is a continuous process. The main challenge in a business which designs, produces, and delivers its own product, is continually understanding customer preferences and customer-base composition, and creating product options which satisfy a variety of needs without compromising operational efficiencies or creating in-feasibility in the supply chain.
This talk will focus on our approach to understanding customer preferences, specifically related to recipe and menu composition, and how we use various engineering techniques to enhance available features. These features are used in a classification model which predicts individual recipe scores during the recipe design process, and can predict overall menu performance based on recipe combinations.
The purpose of this work is to take a data-driven approach to product design, minimize trial and error for our culinary team, and align product design with customer acquisition and growth strategies.

What You’ll Learn:
Attendees will learn about the applications of machine learning to the culinary arts, and more specifically, how to leverage feature engineering and modelling techniques to make data a meaningful part of product design. The presentation will focus on the modelling approach, validating outputs, and immediate business applications.

Delina has over 10 years of experience in data analytics across a variety of domains, including financial services and CPG, and functions, including product management, operations, and marketing. She currently leads a full service data team at HelloFresh Canada, supporting all business functions with engineering, reporting and modelling needs to improve decision making in revenue growth and cost management.

Abstract:
Through this talk, the audience will be introduced to some of the complex problems that impedes the operational efficiency of railway systems. Canadian National Railway(CN) extends from coast to coast across North America. It is important to ensure that the rail cars and railroads are efficiently used with minimal maintenance time. With a network as wide as CN, carrying manual inspections is prohibitively expensive and time consuming. Leveraging computer vision and machine learning to build a novel solution for automated visual inspection would result in higher operational efficiency of the CN.

What You’ll Learn:
Harnessing the power of machine learning for automated visual inspection

Ashley Varghese is a Data Scientist at Canadian National Railway. She works in the Automated Inspection Program overseeing the development and retraining aspects for the inspection of rail cars. She has over 10 years’ of research experience in computer vision and deep learning. Her research papers are published in multiple international conferences and journals. She has previously worked as an AI scientist at Industrial Skyworks and as a Researcher at the TCS Innovation Lab. She holds an MTech in Computer Science from the International Institution of Information Technology Bangalore(IIITB)

Abstract:
We propose a simple method to handle missing features in a graph which is based on feature propagation and is compatible with any GNN model. Our methods outperforms previously proposed approaches in both node-classification and link-prediction tasks, and is able to perform well even when 99% of the features are missing. Moreover, our approach is extremely scalable, running on a dataset with two million nodes in 10 seconds. We theoretically analyze our approach using tools from compressed sensing, showing that it acts as a low pass filter and finding guarantees on how well we can reconstruct the features.

What You’ll Learn:
– Graph Neural Networks are widely used in industry
– In the real world we often do not have all features for all nodes
– We present a simple and scalable approach that works well even if 99% of the features are missing

Emanuele is a Machine Learning Researcher at Twitter, as well as PhD student working on Graph Neural Networks at Imperial College London and supervised by Prof. Michael Bronstein. Before joining Twitter, he was working at Fabula AI, which was then acquired by Twitter in June 2019. Previously, Emanuele completed an MPhil at the University of Cambridge and a BEng at Imperial College London, both in Computer Science.

Abstract:
This talk will discuss projects focused on advancing AI for the common good, including the dramatic creation of the COVID-19 Open Research Dataset (CORD-19) and the broad range of efforts, both inside and outside of the Semantic Scholar project, to garner insights into COVID-19 and its treatment based on this data. It will also highlight key advances in NLP that have enabled this work.

What You’ll Learn:
This talk will discuss projects focused on advancing AI for the common good, including the dramatic creation of the COVID-19 Open Research Dataset (CORD-19) and the broad range of efforts, both inside and outside of the Semantic Scholar project, to garner insights into COVID-19 and its treatment based on this data. It will also highlight key advances in NLP that have enabled this work.

Dr. Oren Etzioni is Chief Executive Officer at AI2. He is Professor Emeritus, University of Washington as of October 2020 and a Venture Partner at the Madrona Venture Group since 2000. His awards include Seattle’s Geek of the Year (2013), and he has founded or co-founded several companies, including Farecast (acquired by Microsoft). He has written over 100 technical papers, as well as commentary on AI for The New York Times, Wired, and Nature. He helped to pioneer meta-search, online comparison shopping, machine reading, and Open Information Extraction.

Abstract:
The key technical problems that online social networks focus on today are detecting policy violating content (e.g., hate speech, misinformation) and ranking content to satisfy their users’ needs. By nature, these problems are somewhat vague and need to handle multi-modal content in many languages, and therefore do not naturally lend themselves to AI techniques based on declarative representations and reasoning. However, the machine learning techniques that are employed also have some drawbacks, such as the fact that it is hard to update their knowledge efficiently or to explain their results. In this talk I will outline a few opportunities where methods from symbolic AI, combined appropriately into the machine learning paradigm, can ultimately have an impact on our goals. In particular, I will describe Neural Databases, a new kind of database system that leverages the strength of NLP transformers to answer database queries over text, thereby freeing us from designing and relying on a database schema.

What You’ll Learn:
New developments on NLP meets data management

Alon Halevy has been a director at Facebook AI since 2019, where he works on Human Value Alignment and on the combination of neural and symbolic techniques for data management. Prior to Facebook, Alon was the CEO of Megagon Labs (2015-2018) and led the Structured Data Research Group at Google Research (2005-2015), where the team developed WebTables and Google Fusion Tables. From 1998 to 2005 he was a professor at the University of Washington, where he founded the database group. Alon is a founder of two startups, Nimble Technology and Transformic Inc. (acquired by Google in 2005). He received his Ph.D in Computer Science from Stanford in 1993. Alon co-authored two books: The Infinite Emotions of Coffee and Principles of Data Integration. He is a Fellow of the ACM and a recipient of the PECASE award and Sloan Fellowship. Together with his co-authors, he received VLDB 10-year best paper awards for the 2008 paper on WebTables and for the 1996 paper on the Information Manifold data integration system. In 2021, he received the Edgar F. Codd SIGMOD Innovations Award.

What You’ll Learn:
Fundamental concepts and methods to understand the predictions, evaluation and trust in ML models

Carlos Guestrin is a Professor in the Computer Science Department at Stanford University. His previous positions include the Amazon Professor of Machine Learning at the Computer Science & Engineering Department of the University of Washington, the Finmeccanica Associate Professor at Carnegie Mellon University, and the Senior Director of Machine Learning and AI at Apple, after the acquisition of Turi, Inc. (formerly GraphLab and Dato) — Carlos co-founded Turi, which developed a platform for developers and data scientist to build and deploy intelligent applications. He is a technical advisor for OctoML.ai. His team also released a number of popular open-source projects, including XGBoost, LIME, Apache TVM, MXNet, Turi Create, GraphLab/PowerGraph, SFrame, and GraphChi.
Carlos received the IJCAI Computers and Thought Award and the Presidential Early Career Award for Scientists and Engineers (PECASE). He is also a recipient of the ONR Young Investigator Award, NSF Career Award, Alfred P. Sloan Fellowship, and IBM Faculty Fellowship, and was named one of the 2008 ‘Brilliant 10’ by Popular Science Magazine. Carlos’ work received awards at a number of conferences and journals, including ACL, AISTATS, ICML, IPSN, JAIR, JWRPM, KDD, NeurIPS, UAI, and VLDB. He is a former member of the Information Sciences and Technology (ISAT) advisory group for DARPA.

Abstract:
As a financial institution, how do you reach out to the right audiences about your products and services in Digital Out of Home (DOOH) world? What is the best content strategy to follow when distributing ads through the network of thousands of digital screens across Canada? In this talk, I will demonstrate a probabilistic modeling approach developed at the CDM and how insights derived from this recommendation model drive decision-making on content selection and content placement.

What You’ll Learn:
To the best of my knowledge, application of Data Science to the Digital Out of Home campaigns is new, and no literature exists on best practices and recommendation system performance evaluation.

Nataliya received her Doctoral Degree in Applied Mathematics from the University of Waterloo in 2010, followed by postdoctoral training at the Neurological Institute in Montreal. Following her postdoctoral assignment, she developed a novel approach to brain tissue classification in early childhood brain MRIs using modern Computer Vision pattern recognition and perceptual image quality models. Nataliya has worked in many industries including biotech, materials science and automotive, and various start-up software companies. Throughout her career, she has applied her expertise in Mathematics to develop numerous models including but not limited to machine learning algorithms and computationally efficient algorithms for model validation. She is the co-inventor of “Bid-Assist”, a strategy for setting up an initial bidding amount to discourage low bidding behaviour, and “AutoVision”, a mobile app that allows automatic taking of pictures of vehicle views and damages recognized by an image classifier. Nataliya paved a new way for Data Science in incentives/rewards industry. She developed predictive analytics tools that help channel leaders maximize the return on investment of their channel incentive programs. In January 2021, Nataliya joined the Cineplex Digital Media as a Senior Data Scientist committed to the development of media content recommendation systems.

Abstract:
MLOps has emerged as a powerful set of tools and strategies for effectively deploying Machine Learning models into production. As an ever-growing number of organizations look to deploy ML models they face an expensive question: how much is this all going to cost? While the research community is chasing the latest and greatest, both data and models are getting bigger and more compute intensive. This trend has led to a ballooning of costs, especially as more models are being deployed into the cloud. In this talk, I will share several effective methods to reduce costs when deploying Machine Learning models. Namely, we will zoom-in on two key areas for cost reduction: models and infrastructure.

What You’ll Learn:
Most of the ML community is chasing the hottest / latest ML models and techniques. The practical implications of effectively deploying models into production are thus often overlooked. In fact, very few practitioners talk about it at all in public so there is little information available. Most of the know-how for deploying ML models into production remains internal to the companies who implement it. My goal with this talk is to distill a practical set of techniques for reducing ML model deployment costs to bring it to the forefront of the community.

George is a Machine Learning Engineer with expertise in bringing Machine Learning technologies to production at scale. In the past, he worked at Indus.ai (acquired by Procore Technologies) designing and building a Machine Learning System for applying Computer Vision to construction analytics. He currently works at Altair Engineering where he’s working on building an open, extensible, scalable, cloud-agnostic MLOps platform to make taking ML to production faster and easier.

Abstract:
In most cases, deep neural networks (DNNs) require a great deal of data. There are approaches, such as zero-shot and few-shot learning, that can produce high quality DNNs with less or no data. However, these approaches still assume a large source dataset or a large secondary dataset to guide the transfer of knowledge. These are not assumptions that hold true when our goal is to model individual humans, who tend to produce much less data. In this talk we present a novel transfer learning method for producing a DNN for modeling the behaviour of a specific individual on an unseen target task, by leveraging a small dataset produced by that same individual on a secondary task. We make use of a specialized transfer learning representation and Monte Carlo Tree Search (MCTS). We demonstrate that our approach outperforms standard transfer learning approaches and other optimization methods on two human modeling domains: financial health and video game design.

What You’ll Learn:
About a novel approach for modelling users

Matthew Guzdial is an Assistant Professor in the Computing Science department of the University of Alberta and a Canada CIFAR AI Chair at the Alberta Machine Intelligence Institute (Amii). He is a recipient of an Early Career Researcher Award from NSERC, a Unity Graduate Fellowship, and two best conference paper awards. His work has been featured in the BBC, WIRED, Popular Science, and Time.

Abstract:
Fireside chat with David Clark, Senior Research Scientist, MIT Computer Science & Artificial Intelligence Lab
As AI evolves, many researchers face a moral dilemma as they watch their work make its way out of the lab and into society in ways they had not imagined or, more importantly, in ways they find objectionable.
As a foundational technology that has reached maturity and is fully embedded in society, the Internet offers powerful lessons about unintended consequences. In this discussion, Dr. David Clark, Senior Research Scientist at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) shares his experiences with the Internet’s transition from lab to market and explains why “the technologists are not in control of the future of technology.

What You’ll Learn:
– How multiple technical, economic, and social requirements shape technology as it transitions from lab to market
– What responsibility, if any, do researchers have for future applications of their work?
– How can researchers have a positive impact?

Natalie Klym is VP Market Development at Radium Cloud as well as an independent researcher and consultant. She has been leading digital technology innovation programs in academic and private institutions for 25 years including at MIT, the Vector Institute, and University of Toronto. She strives for innovation that is open, creative, and responsible.

Clark has been involved with the development of the Internet since the 1970s. He served as Chief Protocol Architect and chaired the Internet Activities Board throughout most of the 80s, and more recently worked on several NSF-sponsored projects on next generation Internet architecture. In his 2019 book, Designing an Internet, Clark looks at how multiple technical, economic, and social requirements shaped and continue to shape the character of the Internet.

Abstract:
Materials development revolutionizes all aspects of our lives. However, the traditional approaches for the design and manufacturing of materials are often unsustainable and resource-intensive. Limited by sparse, decentralized data, and the performance frontier of existing product lines, new materials can take decades to develop. Applications of machine learning and artificial intelligence have been a transformative paradigm for materials science. As materials informatics has matured from a niche area of research into an established discipline, distinct frontiers of this discipline have come into focus, and best practices for applying machine learning to materials are emerging. This talk will discuss case studies of how machine learning accelerates the development of higher-performing products.

What You’ll Learn:
In this talk, I will share very original case studies being performed by AI Materia for the materials/manufacturing industry.

Dr. Maryam Emami is the Chief Executive Officer at AI Materia, a materials informatics platform company. AI Materia technology helps materials companies to discover, manufacture, and design advanced materials in 80% less time compared to traditional approaches.
In addition to her role at AI Materia, Maryam works with materials and manufacturing companies as an advisor around digital transformation and innovation.
She earned her Ph.D. and M.Sc. in Chemical Engineering from McMaster University and holds a B.Sc. in Petroleum Engineering. Her primary interest is the development and validation of physics-informed machine learning methods specific to applications in advanced manufacturing. She is a designated professional engineer, P.Eng

Abstract:
Online social networks such as Facebook and LinkedIn have been an integrated part of people’s everyday life. To improve the user experience and power the products around the social network, Knowledge Graphs (KG) are used as a standard way to extract and organize the knowledge in the social network. This talk focuses on how to build Knowledge Graphs for social networks by developing deep NLP and GNN models, and holistic optimization of Knowledge Graphs and the social network. Building KG for social networks poses two challenges:
1) input data for each member in the social network is noisy, implicit and in multilingual, so a deep understanding of the input data is needed;
2) KG and the social network influence each other via multiple organic feedback loops, so a holistic view on both networks is needed. In this talk, I will share the lessons we learned from tackling the above challenges in the past 9 years on building the Knowledge Graph for the LinkedIn social network. I will present how we use our KG to empower more than 20+ products at LinkedIn with high business impacts.

What You’ll Learn:
You will learn:
1) the overview of the LinkedIn knowledge graph and its business impacts,
2) how does LinkedIn use deep learning techniques, leverage the graph structure, and incorporate the social network feedback to construct the LinkedIn knowledge graph.

Qi He is the Sr. Director of Engineering at LinkedIn, leading a team of 150+ machine learning scientists, software engineers and linguistic specialists to standardize LinkedIn data and build the LinkedIn Knowledge Graph by creating standard entity vocabulary, recognizing entities, building entity relationships, and use this data to serve the entire LinkedIn ecosystem including member engagement and monetization products. His strengths include 1) 15+ years of experience managing and executing large complex AI projects in Knowledge Mining and Management, Recommender Systems, Information Retrieval, Language Processing with big business impact, 2) inventing and driving adoption of the state-of-the-art Deep Learning and Natural Language Processing approaches in industry, 3) building a strong organization and scaling it across geographies.
He is a Member Board of Directors for ACM CIKM and served as General Chair of CIKM 2013 and PC Chair of CIKM 2019. He serves as Associate Editor of IEEE Transactions on Knowledge and Data Engineering (TKDE) and Neurocomputing Journal. He has regularly served on the (senior) program committee of SIGKDD, SIGIR, WWW, CIKM and WSDM for 10+ years. He received the 2008 SIGKDD Best Application Paper Award and the 2020 WSDM 10-year Test of Time Award. He has ~70 publications with 6000+ citations to date. He is an ACM Distinguished Member. He is People of ACM in February, 2021 People of ACM.

Abstract:
Why is it so important to have a strong AI strategy?
Most AI initiatives fail, mainly because there is no strategy, no vision and no framework to support the projects.
L’Oréal Canada has faced several challenges while implementing Data Science and AI practice. The goal of this session is to share L’Oréal Canada experience in the field and how as an organization we think any company should implement data science and AI.
Some of the aspects covered during the session:
•Assessing level of maturity in data valuation and understanding the weaknesses of the organization
•Identify projects with high value, done at L’Oréal Canada
•How to effectively scope a project and avoid POCs?
•How to build a ML team with the shortage of resources in the market?
•In our opinion what are the key points in implementing a data science practice?
This session will be co-presented with the director Data & CRM at L’Oréal Canada and the Partner of Rocket Science Development.
Takeaways for the audience:
A reusable framework in pdf format Use cases idea to drive your AI practice Real examples of projects in Retail and E-commerce Tips and advice based on field experience

What You’ll Learn:
How to assess your current data capacities and maturity (people, technology and processes) and build a viable roadmap for data science in your company.

Senior Global CRM and Customer Experience Strategist with 20 years of success in deploying Data Science, Consumer Centricity and Omni Channel marketing within various local and international organizations.
As CRM & Data Science Director at L’Oréal Canada since 2020, Ludovic leads a team that is solely focused on building better and profitable consumer experiences by bridging data capacities with digital and marketing platforms.

Mohamed Sabri is a results-driven data science and MLOps specialist with 8 years of experience in Machine learning and deep learning, systems design concepts, understanding data architecture concepts, data modeling, project design and project implementation. This Trilingual consultant has a very wide range of project experience for various industries. Mohamed Sabri has been as well a director of data science and AI in a startup in Montréal, with his experience on the field and as a manager Mohamed is capable of supporting all organizations in their AI project implementation. Mohamed is also the author of the book “Data Scientist Pocket Guide” available on Amazon.

Abstract:
Transformer networks have taken the NLP world by storm, powering everything from sentiment analysis to chatbots. However, the sheer size of these networks presents new challenges for deployment, such as how to provide acceptable latency and unit economics. The de-identification tasks Private AI services rely heavily on Transformer networks and involve processing large amounts of data. In this talk, I will go over the challenges we faced and how we managed to improve the latency and throughput of our Transformer networks, allowing our system to process Terabytes of data easily and cost-effectively.

What You’ll Learn:
Tips & tricks for high performance transformer model serving, including the best software stack to use (it makes a huge difference).

Co-founder & CTO of Private AI. He worked on software for Mercedes-Benz and developed the first deep learning algorithms for traffic sign recognition deployed in cars made by one of the most prestigious car manufacturers in the world. He has over 10 years of engineering experience, with code deployed in multi-billion dollar industrial projects. Pieter specializes in ML edge deployment & model optimization for resource-constrained environments.