In this episode of Engineering Choices You Have to Defend, host Nicola Onassis sits down with Ankur Mattoo, technology leader, architect, and AI practitioner, to discuss why the most successful AI initiatives begin years before generative AI ever reaches production.

While helping build the machine learning foundation at Iterable, Ankur faced a challenge common to many fast-growing SaaS companies: enormous amounts of customer data with little consistency. Serving enterprise customers across industries including DoorDash, Spotify, Zillow, and many others, the platform collected highly diverse datasets that were invaluable for marketers—but extremely difficult to transform into scalable machine learning systems.

Rather than rushing to deliver ambitious AI products, Ankur made the strategic decision to invest in foundational infrastructure first. By introducing an incremental product strategy through a feature called Brand Affinity, his team demonstrated immediate business value while quietly building the feature engineering pipelines, machine learning platform, and data foundation that would later support far more advanced AI capabilities.

The conversation explores why strong data architecture, feature stores, and semantic understanding remain essential for successful AI deployments—even in the era of large language models. Ankur explains why organizations that skip foundational investments often struggle to deliver meaningful AI outcomes, while those that balance short-term wins with long-term infrastructure create lasting competitive advantages.

For engineering leaders building AI platforms, this episode offers practical lessons on earning organizational trust, scaling machine learning across complex enterprise environments, and making engineering decisions that continue paying dividends years later.

• Successful AI products are built on strong data foundations rather than AI models alone
• Incremental product wins help secure organizational trust for long-term infrastructure investments
• Diverse customer data requires scalable feature engineering instead of customer-specific machine learning models
• Feature stores create reusable signals that accelerate future AI capabilities
• Enterprise AI success depends on semantic understanding and high-quality data pipelines
• Large language models are only as valuable as the data they can access
• Engineering leaders should balance short-term product delivery with long-term architectural investments
• Building AI infrastructure iteratively reduces technical and organizational risk
• Strong data architecture enables future AI innovation long before it becomes visible to customers
• Curiosity and continuous learning remain essential as AI technologies continue evolving

LinkedIn: linkedin.com/in/ankurmattoo

Website: capgemini.com

Listen Now & Subscribe:

Apple Podcasts, Spotify, Amazon Music, YouTube, iHeartRadio, Captivate, or wherever you get your podcasts.

"Engineering Choices You Have to Defend explores the real technical decisions behind AI systems, enterprise architecture, and scalable software engineering.