Fuzzy.ai - Build smarter software. An AI decision-making API.

And I'm @evanpro's cofounder. Happy to answer any questions you may have. We'll be on here all day!

@walkah Gonna be great!

@novica93 Thanks Nikola! This was the exact situation that @evanpro found ourselves in before starting the company. We just wanted to make our products a little bit smarter, and couldn't find anything that fit in between dumb procedural code and big complicated machine learning systems.

@azizmorsly Thanks for the feedback! We've tried a lot of different options for the onboarding process, and for our target market (developers) the current interactive tutorial seems to perform best for getting users activated. Video seems to stall developers out, and we really want to get people working on their own first agent. But I like the idea of using it as an alternative flow for less technical users -- I'm going to see what we can do there. As for examples: we have some on our blog at https://blog.fuzzy.ai/ but I agree that we could bring that more forward in our pre-signup experience. All good stuff -- appreciate the deep analysis.

@marcboscher Thanks Marc! Appreciate it!

@marcboscher Same here. Great work!

@j_r_wi11iams Hi Jesse, that's exactly right. We've had some users get up and running with just a few hours of work. The main way we differ from some of the other ML-as-a-service offerings is that we try to address the fact that lots of devs are in situations where they know roughly how they want a decision-making system to behave based on past experience (for example, you know what factors tend to indicate a fraudulent transaction if you're using Fuzzy.ai for fraud detection) but don't have the hundreds of thousands of rows of historic data to train a purely statistical model. With Fuzzy.ai, you can start by defining the basic behavior you're looking for with a set of rules, and then as you use the system and provide feedback, it learns and improves automatically over time. In situations where our users know roughly how they want things to work, we've seen them go from signup to working agent in a few hours. Does that make sense? Happy to go in more depth if you'd like!

@mattfogel wow - that's really amazing. Thanks for putting this together, it's long over due. Current ML services are painfully unable to scale ... it makes sense that by limiting the scope of behavior with pre defined rules you can dictate the proper response faster.

We have! We've tried to incorporate the most common in as templates for new agents.

@bryncdn Hey Bryn, that's a great question. Throughout our beta we've been seeing people do some really interesting stuff with Fuzzy.ai, a few of the ones I can talk about are: - Product recommendations - Customer/lead scoring - Matching users in marketplaces - Fraud detection & prediction - Dynamic pricing We expected to see people gravitating toward a few key use cases over time, but it's been interesting to see how diverse it has been. What's also really cool is that once a dev team understands how Fuzzy.ai works, we've been starting to see them come back with more ideas for how to use it.

Thanks @evanpro! Did you guys have a use case in mind when you started building Fuzzy.ai? I'm interested to know if any of the use cases have been surprising.

@bryncdn the original idea came out of @evanpro wanting to build dynamic pricing, and me wanting to build customer scoring at separate companies, so those two have been pretty high up on our radar. Probably the most fun use case that I've had a user bring up is integrating Fuzzy.ai with smart home products to make them even smarter. It's a side project for the person in question, so it hasn't been moving super quickly, but I think it'll make for a really fun blog post once he's done!

@valerioneri Great question! Our customers so far have been comfortable with using a remote API. If we have someone who really wants to use the product on-prem, we would probably be willing to pilot it. (Never say never, right?) But for right now we'd like to minimize the complexity for the majority of developers.

@seanpaley Hey Sean, you can totally do that. All you need to do is take your base set of rules and create multiple agents (one per user or bucket of users in this case). They all start from the same point but can then diverge based on feedback. Make sense?

@mattfogel makes sense, thanks

@lewwwk thanks for the kind words! Would love to know your thoughts and happy to answer any questions that may arise.

@secretmike thanks Mike!

@pedram92d Thanks Pedram! We've already seen some non-developers getting the initial rules set up and then handing things off to their engineering team, which is really exciting, and is a first step to making it possible for just about anyone to do. The challenging part is making it easy for a non-technical user to integrate our API into their product. I'd love to find some good ways to make that happen!

@mattfogel I love the approach if having a middle point handling the connection / translation like how Zapier works. I'm personally super interested in seeing AI leveraged by non-developers like business owners or marketers, we should continue the convo.

@pedram92d absolutely. any time!

@rogerh1991 Thanks Roger!

@jarie_bolander Thanks Jarie. Appreciate the support!

@mdavep thanks Menachem! I'm not aware of anyone using it specifically for auction bids. I know of some who have used it to determine optimal pricing for their products, but not specifically related to auctions. Please give it a try! I'd love to hear your experience.