AI Demo Jam is a new event series organized by Dnipro VC together with Data Phoenix. On March 20, six AI startups presented their products live during the first AI Demo Jam. In this video, you'll see the demo from Future AGI.
Future AGI: build, evaluate, optimize and observe your agentic workflow without human in the loop and ship to prod 10x faster.
Transcript
Nikhil: So I think a lot of guys might be building with AI, right? If whoever is building, it knows that it hallucinates. But for an enterprise use case, that's really bad because you cannot go wrong for 10% of the use cases. That's where we step in. So let's say you're building with these type of… I'm just rushing because that's just three minutes. So let's say you're building with all type of… Yeah, so yeah… We step in and fix it and how does it happen? Let's say you are a company and you want to create an application for, you know, let's say you are in a renovation space and you have a user that is coming on your website and then saying, you know, you want to that with an AI, you want to give him quotes based on his images and everything.
Let's say you want to create this AI. So on our website, on our tool, you can just go and either create a database on that or upload your data. And then just start building. So let's say I have this as my user's input image. And this is a prompt that I gave, and this was a generated image using any of the open source model or your own model. The next step that we do is you can verify all these results, that's where our proprietary technology comes in. You can see if it's working or not. So I defined an eval called, you know, are these enders correct as per the user requirements? And then I can see, I can go back and see, okay, where it failed. So if I see this tender of the image is not correct. So by iterative step, you can create reliable AI apps. So I can do a lot of stuff over here. I can go over here and either run a prompt. So if you see this is a run prompt column. What I essentially do is I choose GPT and run this model and then see all the results. You can do more than that. This is a result where I ran Deepseek GPT and compare the results. And I can see all the traces where they failed, where they passed. And I can observe this in production as well, how they are performing. All these models or agents, how they are performing. So each of the results, where they are failing, I can see them all. I can touch and I can see, you know, what are the results that, what are the decisions that I made, I did in each of the steps. And I can see all the user sessions as well. So all the users what they did, what are the decisions that they did, what are the talks that they had, I can see them all. I can have alerts, I can have pretty charts. Yeah, I think that`s all my time guys. Thank you.
Audience: So, great presentation. I'm just curious on how you arrived at the product and what you're thinking about in terms of pricing?
Nikhil: So, I've been in the AI space since my third startup. Both, I think, first excited, second, acquired. So I've been an AI guy for a long time. So AI fails and it's very bad for enterprises. So then, you know, so AI might be the new software that we think. You're going to have four or five member teams who are creating this. Currently, there's a lot of human in the loop that are involved, as you say, vibe outcomes. So you write a prompt and you see it's out. It's correct. And you deploy it. That's a very wrong way to do it. So right now, it's like 90% of the projects that are created never goes to the end user. So how do you solve it? So that's how we price it. It's like based on usage.So we just have a $50 starting. It's actually free. Anyone can go and start building on it. So the whole platform is actually free in that term, but if you're using a few evals that's based on input tokens, so you can directly go and start building. Thank you.
Dmytro: Thank you.
Comments