Everyone says you need millions to build a real AI company — especially in big data. In this talk I’ll share what actually happens when you don’t have that luxury. I’ll walk through how we built a data-intensive startup with limited funding: choosing where not to spend, prioritizing signal over “perfect” pipelines, and focusing on value instead of hype. We’ll look at the data-science side too — how to get impact before “big data” is really big, how to survive messy real-world data, and how lightweight models plus strong problem framing can beat massive architectures. The goal is simple: show that constraints can be a real advantage, and that it’s possible to build meaningful AI products without burning a fortune.