AI & Machine Learning in the Enterprise

Wayfair may be a home goods retailer, but our customer service representatives are taxi drivers. Our customers are on the street wishing they could catch a ride. In this talk, I’ll describe the Uber of the customer service world: Wayfair’s Algorithmic Contact Assignment (ACA) product. ACA is an adaptive machine learning product that connects customer phone calls and chats to customer service agents. For each customer contact, ACA makes rider/driver-like connections in a hyper-personalized assignment that ensures the customer’s intent, the contact’s complexity, and the value-at-stake of the interaction are handled by the goldilocks agent (the agent with the single best natural proclivities to handle that contact juuuust right). The result is the customer gets what they need faster, happier, and with better resolutions. When customer Sarika got from the parking lot to the soccer field just in time to see her daughter’s first goal because she got off her call with a Wayfair returns specialist in 98 seconds, we knew ACA was perfect!

Peter Wang, CEO at Anaconda

Ying Ying, Bill & Lewis Suit Professor at University of Texas at Austin

Mike Shores, Dr. Director – Data Science at Vista

Sarita Joshi, AI Specialist at Google

Andy Terrel, CTO at Storyfit

In the world of machine learning, ensuring the performance and reliability of models is essential for success. This requires a deep understanding of how the models are behaving and why they are behaving in a certain way. In this presentation, we will discuss the importance of observability in machine learning and how it can help to ensure the performance and reliability of models. We will explore the key metrics that should be monitored, the types of observability tools that are available, and best practices for implementing observability in machine learning workflows.

Let’s talk about mistakes we’ve made while working with data – what we did, what we learned and why we share.

A plain simple question of how we get to keep giving our best, grow up the ladder when there are many many dynamics at play. How to lead with empathy and still win the game. Data team empowerment is a crucial aspect of creating a successful data-driven organization. This results in numerous benefits, including increased productivity, improved data quality, faster insights, and better decision-making. Organizations must ensure that data teams are recognized for their contributions and that their work is valued throughout the organization. Let’s talk a little bit about all this and also how to prioritize data literacy and create a culture that values data-driven decision-making to ensure that data teams are able to collaborate effectively with other departments.

If you’ve ever thought “production is where ML goes to die,” I’m here to propose putting model operations into the hands of the model experts – the data scientists – and making production the value-generating engine it is meant to be. In this session, I deconstruct the challenges many data scientists face and what it looks like to put you at the center of model operations. With purpose-built tools to effortlessly take models from experimentation to production while avoiding operational roadblocks, you will discover how MLOps can be re-imagined to incorporate a data science perspective.

The use of open-source software (OSS) in the enterprise has seen continued growth in recent years, enabling rapid innovation and solution development. Unfortunately, as the use of OSS has increased, so too have software supply chain attacks.

OSS security presents unique challenges that are not typically well-served by traditional IT management and security tools. How do you ensure your open-source pipeline is protected from attacks like the recent Kaseya and Log4j incidents?

In this session, we will discuss:
-Lessons learned from helping secure the OSS pipelines of our customers
-Best practices and common pitfalls to avoid
-The implications of recent regulatory changes related to software supply chain security
-What you can do to prepare for the future

Recent years have seen important advances in the quality of state-of-the-art models; however, this has come at the expense of models becoming less interpretable. In this talk, I will present an overview of the current state of Explainable AI (XAI), within the domain of Natural Language Processing (NLP); the operations and explainability techniques currently available for generating explanations for NLP model predictions, to serve as a resource for model developers in the community.

Kim Martin, Director, Software Engineering at Indeed

Angel Durr, Founder of DataReady DFW and Lecturer at the University of Nevada

Eve Psalti, Senior Director, Artificial Intelligence Engineering at Microsoft

Roja Boina, Chapter Co-Lead at Women in Data

Chris Benevich, Counder & President at Revel Data Incorporated

This talk will outline the current advances in applying AI technologies in medical imaging diagnosis, medical notes, and health risk prediction. The latest AI technologies include transformer, knowledge graph, and explainable AI. This talk summarizes the opportunities and challenges related to AI in health.

Intelligence is more than the human brain. This case study delves into the philosophical underpinnings of how artificial intelligence applied to financial technology problems at the enterprise or platform level reveals new powers of scale. Cognitive science and UX best practices enable effective human-computer-AI interaction at the individual level, which creates an emergent intelligence greater than the sum of its parts once connected to a diverse network of financial institutions.

This presentation provides an overview of the opportunities that arise from changes in human behavior and how to capitalize on them by capitalizing on machine learning and a no-code environment. It will look at how these approaches can help drive revenue growth through predictive interventions, with case studies used to illustrate the effectiveness of this strategy. The presentation will demonstrate how batch-based approaches are no longer suitable for the ever-changing landscape of consumer behavior, and how a continuous machine learning approach, combined with a no-code environment, can enable businesses to stay ahead of the curve and drive profitable growth.

The supply chain is a critical part of the US economy. The freight/trucking industry is the largest component of the US supply chain, both in terms of economic impact ($875B+ in costs and ~11B tons moved in 2021, ATA 2022) and greenhouse gas emissions (420M metric tons equivalent in 2021, US BTS 2022).

In this talk, we will discuss examples of Data Science, Data Engineering, and ML Engineering challenges and corresponding solutions to improve sustainability, efficiency, and profitability for the US trucking market. We will also briefly discuss the challenges in implementing real-time, online solutions at a very fast-growing, fast-moving company with high quality of service expectations.

Machine Learning (ML) is increasingly used to make business-critical decisions across multiple industries. The surge of deep learning has further accelerated this data across multiple data modalities. But unlike traditional software, which has well-defined standards and practices, ML systems lack a systematic way to measure the full spectrum of model quality.

In this talk, we will dive deeper into what ML Data Intelligence means and how it solves the Data Quality problem in Machine Learning, the primary determinant of Model Quality. We’ll be discussing the key principles, techniques, and indicators employed in curating high-quality, error-free datasets for ML and productionizing high-quality models.

Graham Gassle, Head of Data Science, Sales, and Customer Service at Wayfair

Sweta Sinha, Director, Data Science at Ascend Learning

Randi Ludwig, Director, Data Science at Dell Technologies

Alec Coughlin, Client Partner, Financial Services at LivePerson

Companies are frequently faced with large amounts of unstructured text data, like forum comments or product reviews. Important trends can emerge in these datasets, but it can be time-consuming to read through comments, and keyword matching frequently misses critical nuances.

We’ll discuss how we’ve approached this problem at Google using Natural Language Processing, with examples of the approach applied to open datasets.

We’ll explore how this fits into the ML project lifecycle, with examples of common pitfalls. Finally, we’ll highlight how to use this technology as part of a “human in the loop” approach to supercharge your existing team members.

Agile software development practices are great tools to increase the visibility and productivity of any team writing code. However, the data science lifecycle has some particular differences from traditional software development that make it tricky to implement. Randi has led multiple teams of data scientists through the transition of adopting and then thriving using modified Agile methodologies.

She will share tips and tricks for how to get the most value out of Agile in the ways that are most beneficial to the uncertain process of data science. She will also share an adoption roadmap to go from wild west data science to a well-oiled value factory!

Discounting is a pricing technique that can be used to encourage customers (in our case, patients) to pay their bill. If discounting is done right, this can lead to an overall increase in revenue. We applied ML to identify which patients should receive a discount on their bill, but, like many ML projects, it was not as straightforward as it seemed. For example, we realized that the target variable for our model did not fully capture the outcome we are trying to optimize for financial results. By digging into the data to more deeply understand trends in payment behavior, we took steps to build a more robust solution– a journey that is still ongoing!

Running a retail business is hard. At Shopify, we have millions of merchants of varying sizes that face many challenges, including:

– Selling across multiple online channels and managing product metadata, like categories, which is vital for search relevancy and discoverability
-Scaling a business while dealing with an ever-changing online ad space with new regulations

In this talk, Kshetrajna Raghavan will present a holistic view of how Shopify uses interconnected ML systems to solve retail obstacles for Shopify merchants, all while giving these entrepreneurs of all sizes a competitive advantage. Kshetrajna will talk about the tools and platforms Shopify built to support these models, how they evolve continuously, and how they can be applied.

Despite some companies still approaching data science as a pure research field, data science has become a time-tested, value-driving method for companies to solve customer problems. And for some companies that already recognize that data science is a value-creator, the path to unlocking that potential remains unclear.

In this talk, we’ll discuss how embedding data scientists in customer-oriented product teams is a win-win-win recipe for customers, companies, and employees alike. We’ll talk through two specific use cases and how having clearly aligned customer problems, and empowered teams and team members with the right technical skills is a recipe for success.

Hiring is a time-consuming part of a manager’s job. Hiring well is hard and impacts a company’s success. What if managers spent less time hiring really great people? This talk will explore some of the experiments and considerations at Indeed towards finding really great job applicants for any hiring manager.

Companies have to be as flexible as their changing macro-economic environments. Their data and their data teams need to mirror those conditions and built to operate with the constant change of goals and measurables. The right mix of people, processes and systems need to be organized to properly setup companies for success, especially within their data science toolbox. Building the best designed data teams to operate this climate is critical to the company’s and people’s success.

Launching a global and entirely virtual Data Science Hackathon during a pandemic wasn’t easy, but it sparked fantastic innovation and comradery for Visa’s data scientists. If you aspire to launch your own hackathon or any significant event to engage analytic teams, Arya will walk you through all the major milestones you’ll need to plan and execute to make it happen. This presentation will cover topics such as 1) securing executive buy-in, 2) building a budget proposal, 3) proper infrastructure, and 4) what successes you can highlight at the end of the event, among others.

Transfer learning and deep learning unblocked new levels of accuracy for many medical natural language processing tasks. The session shares the current state-of-the-art accuracy on the most widely used healthcare NLP tasks: clinical & biomedical named entity recognition, relation extraction, assertion status (negation) detection, entity resolution (terminologies mapping), and de-identification. You’ll see examples of how Spark NLP enables delivering this level of accuracy in real-world, production systems – adding privacy, tuning, scalability, and the ease of use of getting it all done with a single line of Python code.

Starting up and successfully establishing a data science function or team in organizations of any size is no easy task. From the inherent scientific complexity to differences in how data science teams operate compared to classic data analytics teams to difficulty in finding the right talent, it’s not a journey for the faint of heart. Yet many organizations and leaders have done this and built highly successful teams even amid a once in a lifetime pandemic. This talk will share a set of easy-to-follow Dos & Don’ts and lessons learnt from building up and establishing successful data science teams across multiple industries and companies.

As transformation is pursued, it generally occurs with some amount of friction or resistance, and therefore evolves in highly heterogeneous and disconnected ways across business functions. This presentation will elaborate on the issues faced by traditional firms as they attempt digital transformation in a data-driven world, and describe means to overcome them to achieve outsized and continuous business performance improvement.

Video object tracking is the process of detecting, following, predicting and understanding entities of interest across a video, both spatially and temporally. Conventional video tracking methods are effective in detecting and tracking objects, but typically require an external classification algorithm to identify and discriminate such objects. Similarly, traditional CV methods for image recognition can be too computationally expensive or insufficient for video tracking systems, which need to be lightweight enough to perform at frame rate and robust to random variations in scale, lighting, pose, and occlusion. Current state of the art deep learning methodologies are explored to solve the task of real time video object tracking with contextual understanding. Topics discussed include: DeepSORT, ROLO, Siamese Networks, JDE, GOTURN.

Modern deep learning models are gaining incredible capabilities. As they scale to more complex tasks, they inevitably are becoming larger beasts to tame and train. Training models on larger amounts of data and information contributes to high electricity bills and cloud compute costs, and if we aren’t careful the AI revolution might end up becoming one of the largest carbon emitters today. In this talk, we’ll first learn about bottlenecks slowing down training and inference, and then walk through features in PyTorch that can alleviate these inefficiencies with (mostly) drop-in code replacements.

Mark Moyou – Senior Data Scientist / Solutions Architect at NVIDIA

Connie Yee – Senior Data Scientist at Bloomberg

Jaime Russ – Principal Data Scientist at Ryder Systems

Dr. Man Chon (Kevin) U – VP, Head of Enterprise Data Technology & A.I. at American Express

Charles Irizarry (Moderation) – Principal at Brim & Co, CEO at Phygital

Best practices in data visualization and dashboard design are numerous and sometimes contradictory, but a straightforward method to apply design thinking to creating dashboards is effective and universally applicable. This session will cover the details of design thinking and how it can be applied to dashboard development to create impactful dashboards that meet user needs and provide valuable insights.

Natural Language Processing is a contemporary topic in a variety of different fields. In this talk, we will provide a framework for assessing different machine learning applications generally and specifically on sentiment analysis. We will then go over a couple of case studies on NLP apps for financial and risk management applications, one in deep learning and the other leveraging on-line learning. We will cover their design and architecture then conclude with future extensions, further research and open it up for Q&A.

At Colgate, we have hundreds of models deployed at scale to help our scientists create better products, our marketers stay ahead of trends, and our sales teams optimize their portfolios. These business-facing models play a critical role in supporting decisions and uncovering new opportunities. Choosing the right model for the job is a crucial part of gaining business-acceptance and driving real value. Do we need Inference? Prediction? Prediction with Explainability? Inference with Prediction? This talk will go through some real-world case studies at Colgate where we balance parsimony, prediction, and explainability to deliver models that have tangible impact.

One of the major challenges faced in different enterprises is to generate the value out of the ML models in a short period of time and sustain the investment in the technology to grow it further. The bottlenecks are around the technology and governance behind the data science use cases. The scalability of the solution needs to thought through. The solution is to have right platforms to facilitate high volume feature engineering which uses timely data from different internal and external data sources. Setting up a suitable data ecosystem which is a coherent combination of Data Engineering, Data Science and Visualization technologies can produce really useful data products within an enterprise.

While drafting semitruck is never an option on the highway for obvious safety reasons, the concept of drafting as an analogy for corporate finance does have benefits for the organization. The concept of drafting involves linking vehicles at high speeds to reduce resistance and gain efficiency. If we think of drafting in the corporate finance realm the vehicles become Traditional Financial Models, Advanced Analytic tools and Data Science Machine Learning models. Take a trip with me through discussing drafting these vehicles together to create comprehensive Corporate Finance Fleet Management Solutions.

Joe Salvatore – Chief Risk Officer at Idea Financial

Noelle Silver – Global Partner AI and Analytics at IBM

Sha Edathumparampil – Chief Data Officer at Baptist Health SF

David Smith, VP – Data & Analytics at TheVentureCity

Charles Irizarry – Principal at Brim & Co, CEO at Phygital

Over the last 10 years, Data Science has evolved in many ways, however only few organizations have gained the full competitive advantage from their advanced analytics, despite spending significant amounts of money in their Data Science and Machine Learning (ML) workforce and infrastructure. What could be the reasons? Basically the reasons are two-fold- firstly, most of the companies were rushing to apply data science before they were ready, and the second is the lack of well- skilled data science talents. This talk covers the second part, helping and guiding the Data Scientists to evolve their skill sets with increasing demand of the market. The traditional skill sets of data science are still valuable to build the foundational blocks, but it is important to evolve your skills along with the emerging technologies. As the advancement in technology and tools is trying to bridge the gap of complexity to democratizing data science, one important question that every data scientist should ask themselves- Are you riding the wave in the same direction? If not, you might already be falling behind.

Data teams are an integral part of many organizations nowadays and work not only to generate insights but also to develop tools and products. Many data teams follow agile guidelines, but the exploratory nature of data science creates a hinderance to use agile tools in its traditional form. In this talk I’ll explore aspects of agile and tweak some traditional methods to fit the data science development environment.

Machine Learning and Artificial Intelligence are pushing the boundaries across all the sectors. Health sector is no different. Using ML/AI, computer vision is being revolutionised. This presentation discusses the primary use cases of AI in computer vision for health care, the tool/technologies used, some case studies and major challenges faced.

As NLP has exploded within the world of Data Science and Machine Learning, it is now everywhere in finance as well. Many of us have heard about using NLP to try to outperform the market and predict stock prices. However, NLP is much more versatile than that, and has many other uses as well throughout the finance world. In this talk, I will explore a number of methodologies within NLP, explaining how they are being used in finance to help professionals do their job more efficiently and effectively.

It is  very common for data scientists today to be confronted with the need of processing and developing ML models with very large datasets.

In this chat we will share practical tips that can be applied during the machine learning project lifecycle that  will facilitate architecting  solutions to process large-scale data.

How MLOps is helping REEF to accelerate data science capabilities? In this presentation we will discuss REEF’s Machine Learning Operations Platform and the way the Data Science team work improved as a result of its implementation.

Moody Hadi – Group Manager – New Product Development & Financial Engineering at S&P Global Market Intelligence

Nirmal Budhathoki – Senior Data Scientist at VMware Carbon Black

Charles Irizarry (Moderation) – Principal at Brim & Co, CEO at Phygital