Do you need a big team to get started on your AI idea?

FAQs for AI Zero → One (Episode #5)

Sandeep Uttamchandani
4 min readJun 2, 2022

--

Transcript

Welcome to episode five of Unraveling AI Entrepreneurs. These are basically FAQs from engineers and product managers as you go on your AI journey from zero to one. Today’s question is, what resources do you need to get started on an AI idea? Now, it’s a common misconception that as you think about your AI idea, you really need to get a massive team to build out the product. So you have to make presentations, talk to stakeholders, and get the required support right off the bat. That is a misconception.

What is important is you frame the problem. Um, the earlier episodes where I talked about how to make sure that you have the right success criteria defined. You understand the impact. If this was sold, how impactful it would be for the business. [00:01:00] Those are required, uh, as pre-conditions before you get started. But once you have that, don’t get stuck with getting into, how do I get a team, a massive team together? Instead, your initial goal should really be the cheapest way to verify the feasibility of the problem, and that is key.

You’re not really building out the entire product at this point. You’re taking the riskiest assumptions, the leap of faith assumptions. [00:01:30] Is it possible to build a model? Do we have the data set? Do we have the right precision recall? F1. Whatever is your right success criteria. Can we get there? And just focus on that crux and the resources you need for that initial version.

I like to refer to them as a team trio. You need three to four people in the initial team. On one side, you need someone who is good on the data engineering side. [00:02:00] Understands data and the data pipeline, ETO, the whole process there. Second is the AIML engineer. Someone who can really apply existing algorithms. At this point at an early stage, it’s not about, do we have a problem so unique that we need to invent an algorithm? Most of the time, if you decompose the problem right, and pick the right proxy matrix, you can actually apply existing building blocks.

--

--

Sandeep Uttamchandani

Sharing 20+ years of real-world exec experience leading Data, Analytics, AI & SW Products. O’Reilly book author. Founder AIForEveryone.org. #Mentor #Advise