Schedule

1:00pm ET

ML Highlights from 2021 and lessons for 2022

Oren Etzioni, CEO at Allen Institute for Artificial Intelligence (AI2)

2021 was a year full of advances in machine learning, natural language processing, and computer vision. Inspired by Sebastian Ruder’s blog post, ML and NLP Research Highlights of 2021, this talk will summarize 15 highlights and suggest lessons for 2022 and beyond

Technical Track

Business Track

1:30pm ET

Testing ML models for production

Shivika K Bisen, Lead Data Scientist at PAXAFE

Machine learning models are an integral part of our lives and are now becoming indispensable for decision-making process in many businesses. When ML algorithms make a mistake, it can not only adversely affect the user trust but can also cause loss of businesses and in some sectors – loss of life (health). How do you know that the model you’ve been developing is reliable enough to be deployed in the real world? In this talk, we are going to have a closer look at the Testing ML model for production. Main components of the talk will be :- a) Unit testing b) API Integration testing c) Simulation testing for ML model

Recommendation systems: From A/B testing to deep learning

Uri Goren, Head of Recommendation at Argmax

Recommendation systems got a lot of focus in recent times due to the increase in online shopping. Recommendation always goes hand in hand with measurement and experimentation. In this talk we would cover contextual-bandits, a technique that combines both aspects and bakes machine/deep learning into the process. Contextual bandits are increasingly adopted in the industry, and is being used by recommendation giants such as Netflix, Facebook, Expedia, and many more.

2:00pm ET

Talk by Sanjay Yermalkar

Sanjay Yermalkar, Sr. Director, Data Science Engineering at Anthem

Abstract Coming Soon

Stop Making Data Scientists Do Systems

Emily Curtin, Senior Machine Learning Engineer at Mailchimp

Data Scientists aren’t Systems Engineers, so why do our tools expect them to understand arcane k8s errors? Why do our people systems effectively model them as weird web developers? Many organizations are lacking in a practical understanding of the Data Scientist persona from a UX perspective. By defining what Data Scientists are good at, and more importantly what they’re not good at, we as MLOps professionals and organizational leaders can build on that understanding and let Data Scientists do their best work.

3 Key Takeaways

  • The best tools for Data Scientists are low/no-systems, not low/no-code.
  • Velocity comes from good tooling; quality comes from good incentives.
  • Infrastructure abstraction should be a top priority for MLOps professionals.

2:30pm ET

It's The Data, Stupid! How Improving ML Datasets Is The Best Way To Improve Model Performance

Peter Gao, CEO at Aquarium

When working to improve an ML model, many teams will immediately turn to fancy models or hyperparameter tuning to eke out small performance gains. However, the majority of model improvement can come from holding the model code fixed and properly curating the data it’s trained on! In this talk, Peter discusses why data curation is a key part of model iteration, some common data and model problems, then discusses how to build workflows + team structures to efficiently identify and fix these problems in order to improve your model performance.

Informed Guesser, Minimum Viable Model, Heuristic First: Using ML to solve the Right Problems

Eduardo Bonet, Staff Full Stack Engineer – MLOps at Gitlab

As Machine Learning passes its hype, the industry now enters a more mature scene where ML is not perceived anymore as a magical wand, but as a risky, yet powerful, tool to solve a new set of problems, that requires heavy investments in people and infrastructure. In this product-focused talk, we will be looking at steps we can take to decrease the risk of Machine Learning solution dying on the prototype phase: what types of problems are best fit, ideas on how to handle stakeholder expectations, how to translate Business Metrics into Model Metrics, and how to be more confident if we are solving the right problems.

3:00pm ET

15 min break

3:15pm ET

Panel Discussion: How to put ML successfully intro production

Shivika K Bisen, Lead Data Scientist at PAXAFE
Emily Curtin, Senior Machine Learning Engineer at Mailchimp
Eduardo Bonet, Staff Full Stack Engineer – MLOps at Gitlab
Niko Laskaris, Head of Strategic Projects at Comet

Technical Track

Business Track

4:00pm ET

Talk by Resham Sarkar

Resham Sarkar, Sr Manager – Data Science at Slice

Abstract Coming Soon

External Data: You only own 1% of the data, what about the rest?

Alexander Izydorczyk, Head of Data Science at Coatue Management

Abstract Coming Soon

4:30pm ET

How Feature Stores Enable Operational ML

Kevin Stumpf, Co-Founder and CTO at Tecton

Getting Machine Learning applications into production is hard. When those applications are core to the business and need to run in real-time, the challenge becomes even harder. Feature Stores are designed to solve the data engineering challenges of production ML applications, tackling four key problems:

1. Real-time and streaming data are difficult to incorporate into ML models
2. ML teams are stuck building complex data pipelines
3. Feature engineering is duplicated across the organization
4. Data issues break models in production

Talk by Resham Sarkar
Resham Sarkar, Sr Manager, Data Science at Slice

Abstract coming soon.

5:00pm ET

Building Interactive Machine Learning Demos Fast

Abubakar Abid, Machine Learning Team Lead at Hugging Face

Building machine learning demos is important so that non-technical collaborators and endpoint users (e.g. customers, business teams, quality testers) can provide feedback on model development. However, it can be a time consuming process as it involves front end engineering, design experience, and model deployment. In this presentation, we will talk about an open-source Python package, Gradio, which allows machine learning engineers to quickly generate a visual interface for their ML models entirely in Python. Gradio makes accessing any ML model as easy as opening a URL in your browser. We will provide a technical overview of Gradio and discuss real world use cases in which Gradio has been used to accelerate machine learning workflows.

Talk by Gideon Mendels
Gideon Mendels, Co-Founder and CEO at Comet

Abstract Coming Soon

Schedule

Option 2

1:00pm ET

Introduction/Talk by Gideon

Technical Track

Business Track

1:30pm ET

Testing ML models for production

Shivika K Bisen, Lead Data Scientist at PAXAFE

Machine learning models are an integral part of our lives and are now becoming indispensable for decision-making process in many businesses. When ML algorithms make a mistake, it can not only adversely affect the user trust but can also cause loss of businesses and in some sectors – loss of life (health). How do you know that the model you’ve been developing is reliable enough to be deployed in the real world? In this talk, we are going to have a closer look at the Testing ML model for production. Main components of the talk will be :- a) Unit testing b) API Integration testing c) Simulation testing for ML model

Recommendation systems: From A/B testing to deep learning

Uri Goren, Head of Recommendation at Argmax

Recommendation systems got a lot of focus in recent times due to the increase in online shopping. Recommendation always goes hand in hand with measurement and experimentation. In this talk we would cover contextual-bandits, a technique that combines both aspects and bakes machine/deep learning into the process. Contextual bandits are increasingly adopted in the industry, and is being used by recommendation giants such as Netflix, Facebook, Expedia, and many more.

2:00pm ET

Talk by Sanjay Yermalkar

Sanjay Yermalkar, Sr. Director, Data Science Engineering at Anthem

Abstract Coming Soon

Stop Making Data Scientists Do Systems

Emily Curtin, Senior Machine Learning Engineer at Mailchimp

Data Scientists aren’t Systems Engineers, so why do our tools expect them to understand arcane k8s errors? Why do our people systems effectively model them as weird web developers? Many organizations are lacking in a practical understanding of the Data Scientist persona from a UX perspective. By defining what Data Scientists are good at, and more importantly what they’re not good at, we as MLOps professionals and organizational leaders can build on that understanding and let Data Scientists do their best work.

3 Key Takeaways

  • The best tools for Data Scientists are low/no-systems, not low/no-code.
  • Velocity comes from good tooling; quality comes from good incentives.
  • Infrastructure abstraction should be a top priority for MLOps professionals.

2:30pm ET

It's The Data, Stupid! How Improving ML Datasets Is The Best Way To Improve Model Performance

Peter Gao, CEO at Aquarium

When working to improve an ML model, many teams will immediately turn to fancy models or hyperparameter tuning to eke out small performance gains. However, the majority of model improvement can come from holding the model code fixed and properly curating the data it’s trained on! In this talk, Peter discusses why data curation is a key part of model iteration, some common data and model problems, then discusses how to build workflows + team structures to efficiently identify and fix these problems in order to improve your model performance.

Informed Guesser, Minimum Viable Model, Heuristic First: Using ML to solve the Right Problems

Eduardo Bonet, Staff Full Stack Engineer – MLOps at Gitlab

As Machine Learning passes its hype, the industry now enters a more mature scene where ML is not perceived anymore as a magical wand, but as a risky, yet powerful, tool to solve a new set of problems, that requires heavy investments in people and infrastructure. In this product-focused talk, we will be looking at steps we can take to decrease the risk of Machine Learning solution dying on the prototype phase: what types of problems are best fit, ideas on how to handle stakeholder expectations, how to translate Business Metrics into Model Metrics, and how to be more confident if we are solving the right problems.

3:00pm ET

15 min break

3:15pm ET

Panel Discussion: How to put ML successfully into production

Technical Track

Business Track

4:00pm ET

Talk by Resham Sarkar

Resham Sarkar, Sr Manager – Data Science at Slice

Abstract Coming Soon

What's missing from the Modern Data Stack?

Alexander Izydorczyk, Head of Data Science at Coatue Management

Abstract Coming Soon

4:30pm ET

Talk by Kevin Stumpf

Kevin Stumpf, Co-Founder and CTO at Tecton

Abstract Coming Soon

ML Highlights from 2021 with thanks to Sebastian Ruder

Oren Etzioni, CEO at Allen Institute for Artificial Intelligence (AI2)

2021 was a year full of advances in machine learning, natural language processing, and computer vision. Inspired by Sebastian Ruder’s blog post, ML and NLP Research Highlights of 2021, this talk will summarize 15 highlights and takeaways.

5:00pm ET

Talk by Abubakar Abid

Abubakar Abid, Machine Learning Team Lead at Hugging Face

Abstract Coming Soon

Talk by Saira Kazmi

Saira Kazmi, Senior Director – Enterprise Data Engineering Strategy and AI at CVS Health

Abstract Coming Soon