1Mby1M Virtual Accelerator AI Investor Forum: With Krishnakumar Natarajan, Co-Founder of Mela Ventures (Part 3)

Sramana Mitra: So, tell me, what AI companies have you invested in? Let’s do a few case studies and as you are describing them, talk about why you have invested in those companies.

Krishnakumar Natarajan: I’ll probably talk about cases where all the entrepreneurs present in the webinar can also get a feel of what the problem being solved is.

One of the companies which we invested was based on our thesis that while there’re a lot of AI use cases where the data is critical, the next big wave of AI application is going to be how AI is used on computer imaging.

We have precedence of companies which are attempting autonomous driving, which is again, based on applying AI on computer images.

Sramana Mitra: Yes. That’s a computer vision problem.

Krishnakumar Natarajan: Today, when brands conduct a store audit, they have a merchandiser, either in-house or outsourced, physically visit the store. The merchandiser audits the store to determine how much space their brand occupies – what percentage of that is at eye level, which significantly impacts consumer buying behavior. They also assess competitive activity, which is done manually in most markets. Brand managers then supplement this data with AC Nielsen store audit data to understand how their brands are performing across various locations and geographies.

This company has built an AI model based on computer imaging, integrated with the merchandiser’s smartphone. Linked to the smartphone’s SDK, the merchandiser takes a set of pictures during store visits. Through AI on the edge, they gather data on shelf space and competitive activity.

Brands do not necessarily need this data in real-time, but even at the end of the day, it provides valuable insights into market share and key retail parameters for a specific location. This allows brands to react quickly to competitive moves and predict stockouts. If certain SKUs are low in stock, the system compares this to the planogram and automates the ordering process.

This approach ensures the entire supply chain, especially the last mile, is optimized. It aligns with the vision of creating the “perfect store,” Paul Polman’s concept many brand managers are now trying to implement by leveraging technology. Previously, manual processes validated by AC Nielsen reports were not as effective.

Today, brands like Procter & Gamble and Kimberly-Clark use this technology across markets.

Sramana Mitra: So, this is a merchandiser deployed by FMCG companies like Procter & Gamble and Unilever to conduct store audits.

Krishnakumar Natarajan: That’s right.

Sramana Mitra: Interesting. In what condition did this company come to you? Were they able to show you when you first invested in the company? Were they already showing you the AI that is in action in stores?

Krishnakumar Natarajan: To be honest, no. I think the team which built this used to work for the IBM Visual Commerce business. They had done a lot of work on that, so they just had the idea.

We have now been invested in them for almost five years. Many of the technology and tools available today didn’t even exist at that time. When they started, the accuracy was probably at a 30% level. You found that the ambient lighting in a store made a big difference in terms of accuracy of how you would recognize a product. They had to get into a very different thing of saying, how do you, within the AI model, build in the lighting environment to make the prediction accuracy higher?

Today, they’re running at a 95% plus level of accuracy. The models have been trained over the last three to four years. In a way, this brings us back to our earlier point of discussion. While there’s a lot of talk about AI, it’s not something which you can just work out of the box.

I think there’s a lot of input needed in terms of high quality data, the training approaches you follow, and how do you continuously track the model, whether it’s leading to some hallucination or is there drift in the model?

So, it’s a continuous process of tracking all these metrices.

This segment is part 3 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: With Krishnakumar Natarajan, Co-Founder of Mela Ventures
1 2 3 4