🎧 The AI Investing Strategy of Uber’s 1st Head of Data Science

My guest today is Kevin Novak, the founder and Managing Partner of Rackhouse Ventures, an early-stage venture firm focused on the intersection between AI and the real world. 

Kevin has had a long and respected career in Data Science and AI, serving as Uber’s first head of data science and inventing Uber's dynamic pricing feature. Before Rackhouse, he was the Chief Data Officer of Tala Financial, advised several VC firms, and personally supported over 50 companies as an angel investor.

Today, we discussed all aspects of AI investing, from identifying when data and AI are core to a business model to distinguishing between wrappers and new applications. We also talked about the common characteristics of industries most ripe for AI solutions. In addition, we explored how AI will shape the future of company formation and the venture capital industry itself.

Learn about the strategic application of AI in investment and industry today:

Spotify | YouTube | Apple Podcasts

🗒 Show notes:

[02:01] Kevin's Journey From Working in Academia to Joining the Tech Industry

[05:25] Early Exposure to Big Data, and How Transformative Technologies Like Transformers Are

[15:13] Founding of Rackhouse and Kevin’s Investment Strategy 

[20:57] Detailed Insights into How AI is Reshaping Decision-making Processes.

[30:07] Evaluation of Market Trends and Hype 

[35:00] The Future of AI and Business Models 

[53:28] Advice for Entrepreneurs and Investors

🧠 Key Takeaways

#1 Truly data-driven companies distinguish themselves by understanding and influencing how their customers use data to make decisions. This involves in-depth market and customer research to pinpoint how improvements in data quality or execution can significantly impact procurement decisions, especially in industries burdened with legacy systems and numerous vendors.

#2 As an investor, one must be cautious with industries that are trendy due to data-related trends. For example, skepticism in the Web3 space stems from the fact that high token prices do not guarantee better AI execution. This selective approach avoids markets where data is less likely to be a key decision factor.

#3 In B2B SaaS, the belief has always been that you need one standout feature that's 10x better than everything else. However, an uncommon pattern is that founders often work on two or three medium-value features and stack them together. Kevin found that the same applies to LLMs, especially in co-pilots. There's no real need to migrate everything in your workflow. Good co-pilots naturally integrate with a seamless UX, allowing you to switch contexts effortlessly. It is possible to achieve product-market fit by accumulating a variety of use cases, even if none of them is 10x.

Prefer working with companies that excel in their core vision, which naturally attracts market and business opportunities. This straightforward strategy is particularly effective in environments where adding complexities like AI, extensive marketing, or insider knowledge does not align with the company’s strengths or market effectiveness.

😲You won’t hear this anywhere else

Kevin Novak discusses an investment in a company that is innovating in predictive modeling, particularly for railway safety. He explains:

"In predictive modeling, we face a stark reality: you can't wait for 10,000 trains to derail to collect enough data to prevent future accidents. They only had about 60 photos of cracked wheels from actual derailments. The solution? They’ve developed the ability to synthetically create images of potential defects, which experts can then review and validate for authenticity. Combining these synthetic images with our few real ones, we've created an incredibly effective predictive vision system. This breakthrough in synthetic data technology is opening doors for industries that sparse datasets have traditionally hindered."