Tips on how to build an AI product team: Q&A with Rukmini Iyer

Rukmini Iyer, a Bay Area leader and corporate vice president, heads a dynamic, innovative group focused on sciences and infrastructure to improve the Bing search platform.

The team combines machine learning specialists, game theorists, economists, scientists, and software engineers, each contributing distinct expertise to tackle intricate product challenges. They oversee the algorithmic elements of the search advertising stack, ranging from foundational natural language processing for query and ad comprehension to machine learning of user and advertiser signals and their integration into ad auctions.

We spoke with Iyer to discover how her team operates and to grasp her strategy for developing a successful AI product:

What is the focus of an AI product team?

A common misconception about working on an AI product is that you will be confined to production models and code, with no conferences, cutting‑edge research, or large‑scale models. This could not be further from reality.

The team maintains a balanced blend of product and research. Since AI embedded in advertising is highly sophisticated, it must remain at the forefront of innovation, which is facilitated by research and partnerships with Microsoft Research. Simultaneously, we deploy large‑scale models and machine‑learning data pipelines that must operate within production latency and CPU limits—a substantial software‑engineering hurdle. We conduct both offline and online experiments to rigorously test these models before they are released.

A blend of pragmatism and the pursuit of top‑tier AI characterizes the team. Many are unaware of how swiftly research can be translated into product; with the appropriate team and infrastructure, the process can take only a few months.

Do you look for specific areas of expertise when building a team?

AI differs from other engineering disciplines in that it depends heavily on mathematics and computer science. Breakthroughs arise when a diverse group learns and evolves alongside the product. Experts in computer science, statistics, physics, and those with robust mathematical foundations are valuable assets. Our team includes individuals from varied backgrounds, fostering an engaging learning atmosphere.

The AI landscape evolves rapidly; you need teammates who are passionate about continuous learning.

What are some keys to success in creating a successful AI product?

If you intend to rely on AI for major tasks within your product, you must fundamentally rethink your business model. This involves crafting a clear problem statement, establishing comprehensive metrics and guardrails, and implementing robust experimentation procedures, data pipelines, and analytics.

Equipping employees with the necessary infrastructure and tools is crucial. State‑of‑the‑art AI cannot be built on inadequate infrastructure. Leaders and managers must be involved throughout the entire process, not merely in the initial model development and deployment.

What’s next in AI that you’re keeping tabs on?

Causality, explainability, and a responsible AI framework are among the most important emerging concepts in the field.

We previously trained models on human‑generated data. Today, the data we employ is produced through AI and human interaction, tailored to product and business success metrics, giving it a distinct character.

We are now questioning what it means for AI to learn from itself in real‑world contexts.

If you wish to join Rukmini’s team and work on our AI products, opportunities are available.