Verta MLOps Salon Welcome and Kickoff

Manasi Vartak - Founder & CEO at Verta

Improve Machine Learning Team Productivity with Standardization and Automation - an Introduction to Skelebot

Sherry Wang - Senior Data Scientist at Cars.com

There are many huddles a machine learning engineer/data scientist has to overcome to bring a model from ideation to production, especially on the cloud. How to recreate the development environment easily at a cheap price? Is there a way to enable teams with limited knowledge of container technology to self-deploy docker images? How to version control code as well as artifacts for a ML project? How to encourage teams to continuously create tools to automate repetitive tasks involved in the ML development process? In this talk, the audience will get to see how we solved these problems at cars.com by creating the open source tool skelebot, and be inspired to apply it to solve their own problems as well.

Accelerating ML Workflow with Kubeflow, ModelDB, and Feast

Jaiganesh Prabhakaran - Machine Learning Engineer, Data & Machine Learning at Zulily

A key to Zulily’s success is having a robust ML infrastructure and a wide array of ML solutions that use both structured and unstructured data. This conversation will cover Zulily's ML platform, which is built on top of Kubernetes and uses open-source tools, such as Kubeflow, ModelDB, and Feast, and how the platform helps accelerate ML workflow for ML engineers and data scientists within Zulily’s data science and analytics teams.

How to transform experimental AI projects into successful products

Meeta Dash - VP of Product at Verta

Have you wondered why more than half of the AI initiatives fizzle out in the pilot state? Where are the gaps - business strategy, people, process or tools?

In this talk I will talk about how product managers can help steer the ship in the right direction and increase the success rate of AI initiatives.

- How to align the teams to take a business-first approach to AI initiatives ?
- What is the optimal team structure and process so AI initiatives can adopt agile ?
- Why is a culture of experimentation complemented by the right tools and processes a must ?
- Can you strike a balance between faster product iteration, better UX, predictable releases while reducing business and ethical risks ?

I will also share my personal journey turning a fledgeling AI engineering project into a high value product.

Automation and the need of CI/CD pipeline in machine learning development

Lukas Gubo - MLOps Project Manager at Virufy

Our strategy is to deploy robust but explainable models. To achieve this we are using several open source tools to help us with model development. Many models would remain unused and not brought up to the production stage. Therefore, we count on these tools and are able to make a streamlined process of deploying models.
Target deployment is cloud since with on.prem we are tied to the hardware we have. Cloud provides the best environment for our inference and batch predictions. How do all these components tie together? Let's dive in into more details!

Building capabilities for ML Model Development and Training: Challenges and Best Practices

Meenakshi Sharma - Technical Product Manager at Wayfair

To achieve success with ML, organizations have been investing in Machine learning platforms and operations. The goal is to provide their teams, such as data scientists and engineers, to be able to perform repeatable and scalable model lifecycle management starting from exploration, development, training to the final deployment and monitoring. While there are a wide range of frameworks, tools and technologies available, developing robust, scalable and efficient pipelines still remains complex due to the dynamic nature of the ML workflows. In this session, the speaker will discuss design considerations, challenges and best practices while building platforms for model development and training.

Evolving your ML solutions with collaboration and technology

Orestes Castaneda - Senior Business Intelligence Analyst at CBS Interactive

As businesses and their competitive landscape evolve, more challenging business questions arise. This provides unique opportunities to develop ML applications to enable data-and-intelligence-based decision making. With that, BI and Data Science teams need to quickly adapt and mature their processes to work efficiently and innovate. In this session, I will share process, collaboration and technology components that organizations and teams can use to evolve their solutions portfolio.

Automating AI development for the Edge

Brian Cruz - Head of Core AI at Samba TV

The ability to run Artificial Intelligence algorithms at the edge provides a number of benefits not just limited to better latency, privacy, and cost. In fact, the advent of edge computing has initiated a new paradigm for distributed computing that enables entirely new classes of products and services. Creating new AI-enabled products within this new ecosystem brings with it a number of challenges, one of which is the automation of the development process itself. The development processes for hardware, software, and AI often differ, and this introduces a challenging set of constraints on projects. By combining the time-tested best practices from each of these three areas we can automate many parts of the development process in order to build better products that go to market more quickly.

45 min Panel Discussion

Manasi Vartak - Founder & CEO at Verta
Conrado Miranda - Co-Founder & CTO at Verta
Meenakshi Sharma - Technical Product Manager at Wayfair
Lukas Gubo - MLOps Project Manager at Virufy

Wrap up!