From Data Scientist to ML / AI Product Manager

Insights and tips on how to prepare for a successful transition

As Artificial Intelligence is becoming more and more popular, more companies and teams want to start or increase leveraging it. Because of that, many job positions are appearing or gaining importance in the market. A good example is the figure of Machine Learning / Artificial Intelligence Product Manager.

In my case, I transitioned from a Data Scientist role into a Machine Learning Product Manager role over two years ago. During this time, I have been able to see a constant increase in job offers related to this position, blog posts and talks discussing it, and many people considering a transition or gaining interest in it. I have also been able to confirm my passion for this role and how much I enjoy my day-to-day work, responsibilities, and value I can bring to the team and company.

The role of AI / ML PM is still quite vague and evolves almost as fast as state-of-the-art AI. Although many product teams are becoming relatively autonomous using AI thanks to plug-in solutions and GenAI APIs, I will focus on the role of AI / ML PMs working in core ML teams. These teams are usually formed by Data Scientists, Machine Learning Engineers, and Research Scientists, and together with other roles are involved in solutions where GenAI through an API might not be enough (traditional ML use cases, need of LLMs fine tuning, specific in-house use cases, ML as a service products…). For an illustrative example of such a team, you can check one of my previous posts “Working in a multidisciplinary Machine Learning team to bring value to our users”.

Working in a multidisciplinary Machine Learning team to bring value to our users

In this blog post, we will cover the main skills and knowledge that are needed for this position, how to get there, and learnings and tips based on what worked for me in this transition.

The most important skills for an ML PM

There are many necessary skills and knowledge needed to succeed as an ML / AI PM, but the most important ones can be divided into 4 groups: product strategy, product delivery, influencing, and tech fluency. Let’s deep dive into each group to further understand what each skill set means and how to get them.

Product Strategy

Product strategy is about understanding users and their pains, identifying the right problems and opportunities, and prioritizing them based on quantitative and qualitative evidence.

As a former Data Scientist, for me this meant falling in love with the problem and user pain to solve and not so much with the specific solution, and thinking about where we can bring more value to our users instead of where to apply this cool new AI model. I have found it key to have a clear understanding of OKRs (Objective Key Results) and to care about the final impact of the initiatives (delivering outcomes instead of outputs).

Product Managers need to prioritize tasks and initiatives, so I’ve learned the importance of balancing effort vs. reward for each initiative and ensuring this influences decisions on what and how to build solutions (e.g. considering the project management triangle – scope, quality, time). Initiatives succeed if they are able to tackle the four big product risks: value, usability, feasibility, and business viability.

The most important resources I used to learn about Product Strategy are:

• Good vs bad product manager, by Ben Horowitz.
• The reference book that everyone recommended to me and that I now recommend to any aspiring PM is “Inspired: How to create tech products customers love”, by Marty Cagan.
• Another book and author that helped me get closer to user space and user problems is “Continuous Discovery Habits: Discover Products that Create Customer Value and Business Value”, by Teresa Torres.

Product Delivery

Product Delivery is about being able to manage a team’s initiative to deliver value to the users efficiently.

I started by understanding the product feature phases (discovery, plan, design, implementation, test, launch, and iterations) and what each of them meant for me as a Data Scientist. Then followed with how value can be brought “efficiently”: starting small (through Minimum Viable Products and prototypes), delivering value fast by small steps, and iterations. To ensure initiatives move in the right direction, I have found it also key to continuously measure impact (e.g. through dashboards) and learn from quantitative and qualitative data, adapting next steps with insights and new learnings.

To learn about Product Delivery, I would recommend:

• Some of the previously shared resources (e.g. Inspired book) also cover the importance of MVP, prototyping and agile applied to Product Management. I also wrote a blog post on how to think about MVPs and prototypes in the context of ML initiatives: When ML meets Product — Less is often more.
• Learning about agile and project management (for example through this crash course), and about Jira or the project management tool used by your current company (with videos such as this crash course).

Influencing

Influencing is the ability to gain trust, align with stakeholders and guide the team.

Compared to the Data Scientist’s role, the day-to-day work as a PM changes completely: it is no longer about coding, but about communicating, aligning, and (a lot!) of meetings. Great communication and storytelling become key for this role, especially the ability to explain complex ML topics to non technical people. It becomes also important to keep stakeholders informed, give visibility to the team’s hard work, and ensure alignment and buying on the future direction of the team (proving how it will help tackle the biggest challenges and opportunities, gaining trust). Finally, it is also important to learn how to challenge, say no, act as an umbrella for the team, and sometimes deliver bad results or bad news.

The resources I would recommend for this topic:

• The complete stakeholder mapping guide, Miro
• A must read book for any Data Scientist and also for any ML Product Manager is “Storytelling with data — A Data Visualization Guide for Business Professionals”, by Cole Nussbaumer Knaflic.
• To learn further about how as a Product Manager you can influence and empower the team, “EMPOWERED: Ordinary People, Extraordinary Products”, by Marty Cagan and Chris Jones.

Tech fluency

Tech fluency for an ML / AI PM, means knowledge and sensibility in Machine Learning, Responsible AI, Data in general, MLOPs, and Back End Engineering.

Your Data Science / Machine Learning / Artificial Intelligence background is probably your strongest asset, make sure you leverage it! This knowledge will allow you to talk in the same language as Data Scientists, understand deeply and challenge the projects, have sensibility on what is possible or easy and what isn’t, potential risks, dependencies, edge cases, and limitations.

As you are going to lead products with an impact on users, including responsible AI awareness becomes paramount. Risks related to not taking this into account include ethical dilemmas, company reputation, and legal issues (e.g. specific EU laws like GDPR or AI Act). In my case, I started with the course Practical Data Ethics, from Fast.ai.

General data fluency is also necessary (probably you have it covered too): analytical thinking, being curious about data, understanding where data is stored, how to access it, importance of historical data… On top of that it is also important to kow how to measure impact, the relationship with business metrics and OKRs, and experimentation (a/b testing).

As your ML models will probably need to be deployed in order to reach a final impact on users, you might work with Machine Learning Engineers within the team (or skilled DS with model deployment knowledge). You’ll need to gain sensibility about MLOPs: what it means to put a model in production, monitor it, and maintain it. In deeplearning.ai, you can find a great course on MLOPs (Machine Learning Engineering for Production Specialization).

Finally, it can happen that your team also has Back End Engineers (usually dealing with the integration of the deployed model with the rest of the platform). In my case, this was the technical field that was further away from my expertise, so I had to invest some time learning and gaining sensibility about BE. In many companies, the technical interview for PM includes some BE related questions. Make sure to get an overview of several engineering topics such as: CICD, staging vs production environments, Monolith vs MicroServices architectures (and PROs and CONTs of each setup), Pull Requests, APIs, event driven architectures….

Wrapping up and final tips

We have covered the 4 most important knowledge areas for an ML / AI PM (product strategy, product delivery, influencing and tech fluency), why they are important, and some ideas on resources that can help you achieve them.

Just like in any career progress, I found it key to define a plan, and share my short and mid term desires and expectations with managers and colleagues. Through this, I was able to transition into a PM role in the same company where I was working as a Data Scientist. This made the transition much easier: I already knew the business, product, tech, ways of working, colleagues… I also looked for mentors and colleagues within the company to whom I could ask questions, learn specific topics from and even practice for the PM interviews.

To prepare for the interviews, I focused on changing my mindset: developing vs thinking whether to build something or not, whether to launch something or not. I found out BUS (Business, User, Solution) is a great way to structure responses during interviews and enforce this new mindset there.

What I shared in this blog post can look like a lot, but it really is much easier than learning python or understanding how back-propagation works. If you are still unsure whether this role is for you or not, know that you can always give it a try, experiment, and decide to go back to your previous role. Or maybe, who knows, you end up loving being an ML / AI PM just like I do!

From Data Scientist to ML / AI Product Manager was originally published in Towards Data Science on Medium, where people are continuing the conversation by highlighting and responding to this story.