# Machine Learning Care

Matthew Dombrowski | MLOps Podcast | Episode 142 | 16:15
Hosted by Mihail Eric

Source: https://www.youtube.com/watch?v=vgEdlOxmzRo
Channel: MLOps Community, now AAIF Live (https://www.youtube.com/@AAIFLive-x1r). Summarised by MLOps Talks.
Page: https://mlopstalks.com/talks/machine-learning-care
Published: 2023-02-07
Tags: careers, platform-teams

## TL;DR
- Machine learning products need people handling customer work, implementation, project management, product decisions, and go-to-market work alongside engineering and science.
- Matthew Dombrowski moved from computer science and network security into solutions consulting, solutions architecture, machine learning implementation, and product management.
- ML product managers often own internal data or platform products, so they need technical and domain knowledge plus strong internal communication and stakeholder alignment.

## Summary
Matthew Dombrowski describes the range of work involved in building and shipping machine learning products. He argues that the field is often reduced to data scientists and ML engineers, even though larger companies need people for onboarding, implementation, professional services, customer-facing work, project management, sales support, and product management. He traces his own path from computer science and the Air Force through network security, solutions consulting, solutions architecture, machine learning implementations, and product management. Mihail Eric then asks how machine learning product management differs from more familiar software product roles. Matthew describes ML product managers as often working on data or platform products for internal teams. Their work includes setting roadmaps, defining platform capabilities, and helping other teams move faster. He says these roles require more technical and domain knowledge than some other PM roles, along with communication and the ability to align internal stakeholders and secure resources.

## Key ideas
### Machine learning products need more roles than engineering and science
[00:00](https://www.youtube.com/watch?v=vgEdlOxmzRo&t=0s)
Matthew says people often think a machine learning career requires becoming a data scientist or ML engineer. Building and launching a successful product also involves go-to-market work, customer-facing work, internal project management, and other coordination. Larger companies have many moving parts and may need a sizable team to ship something. He invites people who do not want to write Python for the rest of their careers to consider where they can add value in the wider ML ecosystem.

### Customer onboarding and implementation are real ML product work
[04:27](https://www.youtube.com/watch?v=vgEdlOxmzRo&t=267s)
Matthew began at a mobile analytics startup as a solutions consultant. He helped customers get started with the product and become successful. He describes this kind of onboarding and implementation consulting as a role that can provide white-glove support for ML products and other SaaS products. He later became a solutions architect, handling more difficult or customised implementations, building integrations with customers, and doing post-sale consulting and professional services work.

### Solutions architecture can connect technical work with sales and adoption
[05:34](https://www.youtube.com/watch?v=vgEdlOxmzRo&t=334s)
Matthew explains that the solutions architect title means different things across companies. At AWS, the role was similar to a sales engineer: he worked with sales teams to provide technical guidance, answer customer questions, and overcome technical hurdles in the sales process. After a sale, other consultants might help with onboarding and implementation. AWS used a different model because its services had metered billing, so solutions architects were paired with customers in different ways depending on customer size and organisational structure.

### Matthew's path into ML came through implementation and domain work
[08:27](https://www.youtube.com/watch?v=vgEdlOxmzRo&t=507s)
Matthew moved from solutions architecture into a product manager role on Amazon's Alexa team, where he supported internal customers building models for Alexa use cases. Before people commonly used the term machine learning, he had worked on predictive analytics such as predicting mobile-app churn and responses to push notifications. At AWS, he worked on machine learning implementations for financial services, insurance, and biotechnology customers, using AWS storage and compute services before higher-level services such as SageMaker existed. At Alexa, he supported speech recognition and natural language understanding teams with their infrastructure problems.

### Platform PMs own products used by other teams
[12:48](https://www.youtube.com/watch?v=vgEdlOxmzRo&t=768s)
Matthew describes many ML product roles as platform PM roles. A PM might own internal data products, data sold to other teams or external customers, or the feature set and roadmap of an internal platform. He gives Uber's Michelangelo as an example of this kind of platform. The customers are often internal teams that need platform capabilities to move faster, although some platforms may later support external customers. This work treats data and infrastructure as products with customers, value, and roadmaps.

### ML product managers need technical depth and internal alignment
[15:12](https://www.youtube.com/watch?v=vgEdlOxmzRo&t=912s)
Matthew says ML and platform PM roles usually require more technical inclination than some other PM roles. Domain expertise in machine learning, cloud, or data can also be necessary, and he says some of these roles would be difficult to enter without it. Internal stakeholder alignment and communication are major parts of the job. Product managers are often expected to be strong communicators who can sell ideas internally and make the case for resources.

## Notable quotes
- Matthew Dombrowski: "There's a lot of other opportunity out there besides just those roles especially like big companies too." (01:38)
- Matthew Dombrowski: "There's probably someone that is there to help kind of provide White Glove service to help customers become successful." (05:05)
- Matthew Dombrowski: "I wanted to have more ownership over the engineering and product development process." (11:48)
- Matthew Dombrowski: "A lot of the ML type or ML and PM roles are kind of like platform PM roles." (13:37)
- Matthew Dombrowski: "Companies expect PMs to be the best communicator in the room." (15:36)

## Tools & references mentioned
- MLOps Community
- AWS
- Amazon
- Alexa
- Air Force
- SageMaker
- Michelangelo
- Uber

## Who should watch
- You are considering an ML career but do not want a role focused entirely on writing models or data pipelines.
- You work in customer implementation, solutions architecture, consulting, or sales engineering and want to understand how those skills transfer into ML.
- You are moving into an ML platform or data product role and need a clear view of the technical and stakeholder work involved.

## Related talks

- [I Am Once Again Asking "What is MLOps?"](https://mlopstalks.com/talks/i-am-once-again-asking-what-is-mlops) (Oleksandr Stasyk, Synthesia, 1:06:22)
- [Reinvent Yourself and Be Curious](https://mlopstalks.com/talks/reinvent-yourself-and-be-curious) (Stefano Bosisio, Synthesia, 57:16)
- [The Journey from Data Scientist to MLOps Engineer](https://mlopstalks.com/talks/the-journey-from-data-scientist-to-mlops-engineer) (Ale Solano, 41:30)
- [Machine Learning Operations: What Is It and Why Do We Need It?](https://mlopstalks.com/talks/machine-learning-operations-what-is-it-and-why-do-we-need-it) (Niklas Kühl, IBM and Karlsruhe Institute of Technology (KIT), 58:47)
- [MLOps: A Leader's Perspective](https://mlopstalks.com/talks/mlops-a-leaders-perspective) (Stephen Galsworthy, Quby, 54:08)