Data Science vs. AI Product Management: Tech Careers

In this interview, we delve into AI product management with Juhi Parekh an experienced AI product lead. Juhi shares her career journey, sheds light on the key differences between data science and product management, and offers valuable tips for anyone interested in AI.

Data Science to Product Management

Jean: You started in data science, but now you're a product manager. What made product management interesting, especially with your data background?

Juhi: Here’s the story. An internship at SocialCops using data for social good sparked my interest in product management. Seeing how I could improve their core product with user experience (UX) made me realize product management leveraged both my data skills and UX design strengths. That's when I decided to focus on data for two years before transitioning to product management.

Choosing the first role in tech

Jean: A common question for recent grads is often about choosing the right path. What are your thoughts on choosing the right first role in tech?

Juhi: It can be tough for companies to hire someone fresh out of school for a product manager (PM) role. They often seek someone with experience and at least one core skill, like data analysis or engineering.

Jean: Agreed! I've mentored many computer science students who worry they might not love coding forever. But the reality is that you don't have to pick one path for life. You can try software engineering, and if it's not a good fit, you can explore other areas in tech. Having that engineering background can be a big advantage in many roles.

Juhi: Exactly. As a PM, you move into a more generalist role, touching on many different areas. Having a strong foundation in another field, like engineering, is often a plus.

Data vs Product

Jean: You've worked as both a data scientist and a product manager. Can you break down the key differences between the two roles?

Juhi: Absolutely! They're quite different from my experience at Ola. In my first year, I built a location intelligence product while also acting as a product manager (PM).

Data scientists spend most of their time focused on a specific part of the product, like:

• Cleaning and preparing data
• Building models
• Checking how accurate the models are
• Creating ways to measure success (metrics)
• Improving the models over time (retraining)

Product managers, on the other hand, take a more holistic view. They look at the entire product lifecycle, asking questions like:

• Why are we building this product? What problem are we solving?
• Who are we solving this problem for? (target users)
• How will the product actually solve the problem?

To answer these questions, PMs interact with many different teams:

• Talk to users to understand their needs.
• Review customer complaints to identify problems.
• Quantify the problem (how big is it?).
• Decide which problems are most important to address first (prioritization).
• Get buy-in from others to launch the product.
• Figure out the best way to get the product in front of users (go-to-market strategy).
• Define how success will be measured (metrics).

As you can see, data science is more focused on the technical aspects of building a product, while product management takes a broader view, considering everything from user needs to business goals.

Key Skills for SWE

Jean: From your experience as a product manager (PM), what are some practical tips for software engineers to collaborate effectively with PMs, especially on AI projects?

Juhi: The best engineers I've worked with share some key qualities:

• They care about the "why." They understand the problem we're trying to solve and the purpose behind the AI project.
• They think ahead. They consider future uses for the AI product during the design phase, making it adaptable for new situations.
• They ask questions. Don't be afraid to ask for clarification or challenge assumptions – this helps ensure everyone's on the same page.
• They collaborate. Work openly with the PM to find solutions.
• They provide reasonable pushback. Sometimes, you might need to disagree with a PM's idea, but do it respectfully and constructively.

PMs manage multiple stakeholders – designers, engineers, data scientists, and more. It's a balancing act. Think of yourselves as partners with a shared goal: building the best possible product.

AI Product Life Cycle

Jean: Can you walk us through the typical lifecycle of an AI product?

Juhi: There are two main types of AI/machine learning projects:

1. Finding the right problem to bring a new technology to market: For instance, Apple's augmented reality (AR) could enhance map navigation, but it also has potential uses across various industries.
2. Leveraging AI as one of the solves for an existing problem: For example, at Samsung, I developed a visual search tool that identifies and labels medications by pointing your phone camera at a medicine bottle, seamlessly integrating the information into your health app.

Here's the AI product lifecycle broken down:

I divide an AI product lifecycle into 3 buckets:

• AI stack - focused on developing and improving AI algorithms and models
• Production stack - focused on infrastructure and systems to deploy and serve the AI model
• Real world applications - the deployed AI model applied on real-world applications and services

Simplifying it, the process is:

1. Define the Problem & Success Metrics: What problem are we solving, and how will we measure success? This might involve business metrics (like profit) or user metrics (like how many people use the feature).
2. Data Acquisition, Preprocessing & Training: We need data to train the AI model. This data needs to be cleaned and prepared before training can begin. As the model is trained, we might discover we need more data of a certain type.
3. Building & Refining the Model: The AI model is built and trained on the data. As the model sees more data, it gets better at its job.
4. User Experience (UX) and design: We conduct user studies to determine how well people understand and use the feature. This might lead to tweaks in the design or functionality.
5. Deployment: Finally, the product is launched and made available to users!

As you can see, the AI product life cycle involves many steps, from identifying the opportunity to user testing and finally launching the product.

6. Key to Success in AI

Jean: What are some essential qualities or skills for success in the field of AI product management?

Juhi: There are two main things:

1. Staying on Top of Trends: The world of AI is constantly evolving. Successful AI product managers need to be aware of the latest advancements and research.
2. Understanding Research: Being able to read and understand research papers is crucial. This allows you to stay informed about new techniques and potential applications of AI.

Summary

• While product manager roles often require experience, building a foundation in engineering can be a strong advantage.
• Data scientists focus on technical aspects like building models, while product managers take a holistic view, considering user needs and business goals.
• Effective software engineers for AI projects understand the "why," think ahead, ask questions, collaborate openly, and provide reasonable pushback.
• There are two main types of AI projects: finding problems for new technology and leveraging AI for existing issues.
• The AI product lifecycle involves defining problems, data acquisition, model building, refining, UX design, deployment, and continuous improvement.
• Staying on top of trends and understanding research papers is crucial for staying informed about new techniques and potential applications of AI.