InMobi is one of the largest independent mobile advertising and consumer internet platforms globally, with products spanning intelligent shopping, lock-screen content, and hyperlocal weather. They were betting big on generative AI across their product suite. The bottleneck wasn’t engineering. It was product.
The challenge
InMobi’s product managers were expected to add generative AI features into their products, but most of them didn’t have the technical depth to make good decisions about what to build, how to build it, or where AI would actually add value versus where it would just add complexity.
This wasn’t about learning to write prompts. These PMs needed to understand the technology well enough to spec AI features, have informed conversations with their engineering teams, and make sound calls on when fine-tuning makes sense versus when a simpler approach works. Without that understanding, there was a real risk of shipping AI features that looked impressive in demos but fell apart in production.
What we did
We ran an 8-week hands-on program for 40+ product managers. The focus was practical: not just understanding AI concepts, but actually building and prototyping AI features relevant to InMobi’s products.
The curriculum was designed around real problems the teams were facing. One product team, for example, was using vision-based transformers to identify blank AI-generated images. These images were being shipped as lock-screen graphics, and the model was producing too many false negatives, letting blank images slip through. We worked with the team on fine-tuning approaches that significantly reduced the error rate. That kind of hands-on problem-solving, applied to their actual products, was the core of the program.
By the end of the 8 weeks, PMs weren’t just talking about AI. They were speccing features with a clear understanding of what the models could and couldn’t do, and having much sharper conversations with their engineering counterparts.
The result
Product managers started actively building AI features into InMobi’s products. The blank-image detection fix alone was recognized by the VP of Product as a meaningful improvement to their content pipeline.
More broadly, the program closed the gap between product and engineering on AI. PMs went from treating generative AI as a buzzword to treating it as a set of specific capabilities with specific trade-offs, exactly the kind of understanding you need to ship AI features that actually work.
