How to build inclusive teams in data science and AI

Data science is central to great services in tech.

It offers cutting-edge disciplines and transferable skills that are supercharging new ways of doing business better.

In this piece, Mitra Goswami, Senior Director of Data Science & Machine Learning at PagerDuty, looks at the lessons she’s learned on her journey, exploring how to build inclusive AI products as a female leader in data science. 

Mitra, a doctor in Astrophysics from Northwestern University, has spent the last three years helping to accelerate PagerDuty’s AI & Machine Learning tools, building scalable data-centric teams and cost-effective data engineering solutions to drive innovation.

Many routes in

I received my PhD in Computational Astrophysics, following a Master’s in physics. I was led there by pure curiosity. I used to work with C++, coding, then understanding the physics, and then applying it to model observed star cluster systems. I looked at data from telescopes and applied it to model these systems, in accordance with the laws of physics. This is how my data science journey started – with a foundation of maths, physics and computer science.

One thing to stress is that analytical skills and qualifications are key but aren’t used in isolation. Data science mavens must learn to manage complex maths and models, and how business problems and people work, then integrate them into the same, multi-use mental toolkit.

Understanding the problems and needs of stakeholders really is half the battle – it’s the essential context that creates the conditions for better problem solving and model-making that will solve customer problems.

For example, I brought in expertise and public outreach capabilities from my mentorship experience, and outreach experience with my Institute’s observatory and planetarium. This helped me understand and communicate with my business stakeholders – non tech stakeholders – better. Always ask ‘what is the business problem? How can data help?’

Taking advantage of opportunities

To build a good foundation in data science, maths and computer science are critical. Maths is required to understand how the algorithms work, and computer science for coding and implementing them. In that way, it is easier for young people in STEM to enter the data science and AI fields.

During my Ph.D., in addition to gaining more profound knowledge in the field of astrophysics, I was trained to be curious and ask questions that might help me to solve the research problem at hand. My suggestion to young people would be to always ask questions, even if no one else is asking. That’s how you learn.

It’s also important to remain flexible, especially in data science. For example, maybe I envisioned a solution with high accuracy; however, the business stakeholders prioritised a solution that was less accurate but with better reach. It is essential to tell a story, explain your thought process, and understand your stakeholder’s context. Then, one can build something unique together, championing the customer.

There is a dearth of women in senior leadership positions in the tech industry. It’s critical to find a mentor, people who have seen and faced similar challenges, especially for women. If not within the organisation, one should reach out to wider networks, join mentorship organisations, and build skills by learning from others in the same field.

There are also many social constructs. Women tend not to talk about themselves as much as men – though it’s hard to generalise. I would say learn to take a compliment, it is important, and equally important to understand feedback. Ask for feedback from colleagues and mentors to continuously improve.

Managing data science teams

The best advice I can offer to a data scientist is how critical it is to tell stories about your work from day one. If you’re dependent on someone else to tell that story, it might not be very effective and compelling. Data scientists need to explain the story of what the data is solving, and the journey to solve the problem with the tools you have available.

Uniquely, data scientists have the opportunity to talk to multiple people across teams, share data insights and paint a story for product, engineering, legal, and marketing teams – et cetera. What you find is that data science becomes a connective tissue between stakeholders. In tech, other engineers work in a team and build features for it. A data scientist builds services for many teams with wildly different needs.

Stay on top of innovation. Read widely and carefully to understand where disruption will strike. I block off Friday afternoons for research. I stay up to date with best practices and new solutions. Conferences are a key way to understand where you are in the field, what others are doing, and what you need to do better.

Building better products

Inclusivity is really important in data science and in its teams. It’s critical to have different skills and diversity. It helps everyone improve and learn from each other. With only one set of skills, it’s hard to evolve. Data science needs diversity – and it’s a great place to work.

Doing good with data is one of the biggest challenges/opportunities in tech. For example, many cities use data to map how to distribute police or medical help. Then there is weather forecasting, or carbon emission models – so many applications for data science and AI to drive impact for good.

Stamping out bias and making better models and products helps society holistically. We must be aware of what data and rules we’re training the models on – returning us to ‘asking the right questions’.