Artificial Intelligence startup in the plumbing business

[satty]

Howdy folks I'm the co founder of a bay area based startup called Drishyam.ai. We are building a platform that lets users take pictures of faucets & other plumbing products. Our software will identify the make and model and parts associated with it. Love to get back some feedback on the idea.

Here's a link to a chatbot we built:

Here's an explainer video about how it works:

Check us out @ : www.drishyam.ai

-Satty

 

[FullySprinklered]

As long as HJ holds up, we wont be needing your services.

 

[Sponsor]

 

[satty]

As long as HJ holds up, we wont be needing your services.

Sorry whats HJ?

 

[hj]

IF you need the manufacturer's name, you have already eliminated 90% of the people who could use your service, and if they do have it, they may not need you anyway. What do you do about the 1,000s of faucets that are NOT on the Internet for whatever reason?

 

[IHateUserNames]

Satish:

Consumer here. Not a vendor. Also done startups. So, please take with a grain of salt. You came here for advice, but maybe not my kind! lol.

Looks like you are pushing it as a vendor app.

But 1) I really don't do vendor apps anymore. Too much clutter and tracking for a once in a year or decade need. I might email the photo to them or upload it as part of a support request form. My stove has a plainly visible tag. Kenmore's look alike from the front for the most part. and stove or dishwasher, the model number changes somewhere underneath sometimes because one has steel inside and another plastic

HOWEVER... 2) It could prove useful as a direct to consumer app. If I don't know the make or model, I don't know who to call anyway. If it seems painful or obscure for a part, I might be very tempted to replace the faucet given an excuse to upgrade. If I can ID what I need I've given myself an excuse not to spend money. Either answer works for me. Auto parts. OMG. That would be useful.

And 3) I'm also guessing that you are expecting to use consumer pictures to help build and populate your database. So where "we" can identify a product or come close, it will help add new and categorize old. (the "is it 'this' page"?) Direct to the consumer will get you more images.

https://www.nvidia.com/en-us/deep-learning-ai/startups/

Nice. Good luck. Always nice to spend OPM or resources.

LOL. Came for a sump pit, left talking about a pic/ai pit.. Back to plumbing....

 

[satty]

IF you need the manufacturer's name, you have already eliminated 90% of the people who could use your service, and if they do have it, they may not need you anyway. What do you do about the 1,000s of faucets that are NOT on the Internet for whatever reason?

HJ,
We do make and model. so its Logo + Model we identify. Then would it be useful? Its a process of elimination. With the combination of Logo & model we will identify it as one of the top 20 models in a couple of steps. Others we dont care. Because its a long tail. The 80/20 rule If its a no name chinese manufacturer then you are out of luck anyways right?

 

[satty]

Satish:

Consumer here. Not a vendor. Also done startups. So, please take with a grain of salt. You came here for advice, but maybe not my kind! lol.

Looks like you are pushing it as a vendor app.

But 1) I really don't do vendor apps anymore. Too much clutter and tracking for a once in a year or decade need. I might email the photo to them or upload it as part of a support request form. My stove has a plainly visible tag. Kenmore's look alike from the front for the most part. and stove or dishwasher, the model number changes somewhere underneath sometimes because one has steel inside and another plastic

HOWEVER... 2) It could prove useful as a direct to consumer app. If I don't know the make or model, I don't know who to call anyway. If it seems painful or obscure for a part, I might be very tempted to replace the faucet given an excuse to upgrade. If I can ID what I need I've given myself an excuse not to spend money. Either answer works for me. Auto parts. OMG. That would be useful.

And 3) I'm also guessing that you are expecting to use consumer pictures to help build and populate your database. So where "we" can identify a product or come close, it will help add new and categorize old. (the "is it 'this' page"?) Direct to the consumer will get you more images.

https://www.nvidia.com/en-us/deep-learning-ai/startups/

Nice. Good luck. Always nice to spend OPM or resources.

LOL. Came for a sump pit, left talking about a pic/ai pit.. Back to plumbing....

IHUN,
Thxs for your feedback. Appreciate it. Here's a better option than App download. Chatbots running on Vendor website. No download. Just upload your picture. We will identify, troubleshoot & order parts. Never ever call into support. Thats our promise. The demo video is a real chatbot we built for the manufacturer. We do faucets parts. We could auto parts as well. Thats next.
Its funny you mention the inception program. My startup is part of the inception program . Love to get detailed feedback from you.

-Satty

 

[satty]

IF you need the manufacturer's name, you have already eliminated 90% of the people who could use your service, and if they do have it, they may not need you anyway. What do you do about the 1,000s of faucets that are NOT on the Internet for whatever reason?

HJ,
We can build the solution for any number of faucets in the world. We just need pictures. About 200/Model. Then we can identify with about 90-95% accuracy.
-Satty

 

[IHateUserNames]

@satty I checked out your web site, I knew it was part of Nvdia

 

[Asktom]

I used to sell stems on line and HJ is right, people who know the brand probably won't need you.

The best starting points to identify a stem are it's length and the number of splines on the broach. But, for companies that no longer exist you are going to find that most of the illustrations you need are copyrighted by your competitors. There are also weird spots such as manufactured stems looking different and having broaches with a spline count made to fit available handles when the originals are no longer to be had. Ceramic stems will make you pull your hair, there are a bazillion that look very similar, many of which are not to be had. I used to just shake my head when someone wanted a cartridge for a faucet they bought at Pottery Barn. Also, if you think that consumers can look at two stems that are obviously different to the trained eye and see that they are not the same you have overrated your fellow human beings. This means the return rate will be high. If you overcome all this there is the reality that most customers will be buying one or two items. This is just a heads up, I am no longer in the biz and wouldn't be a competitor.

 

[satty]

I used to sell stems on line and HJ is right, people who know the brand probably won't need you.

The best starting points to identify a stem are it's length and the number of splines on the broach. But, for companies that no longer exist you are going to find that most of the illustrations you need are copyrighted by your competitors. There are also weird spots such as manufactured stems looking different and having broaches with a spline count made to fit available handles when the originals are no longer to be had. Ceramic stems will make you pull your hair, there are a bazillion that look very similar, many of which are not to be had. I used to just shake my head when someone wanted a cartridge for a faucet they bought at Pottery Barn. Also, if you think that consumers can look at two stems that are obviously different to the trained eye and see that they are not the same you have overrated your fellow human beings. This means the return rate will be high. If you overcome all this there is the reality that most customers will be buying one or two items. This is just a heads up, I am no longer in the biz and wouldn't be a competitor.

Thanks for your feedback Tom. Really appreciate it!!! Here's what I'm thinking. We will start with identifying the make and model without peeking under the hood for parts. We will do parts explosion based on the model. We will cater to the top 15 manufacturers & their brands in business. Thats it Or at least for now. We will start with recognizing say 20% of the models for now and then build on top of it. For these 20% We will identify the make and model (Logo + shape) with a 90- 95% accuracy. All the users have to do is a take a picture. We will identify the make and model & the parts. We will see how the market reacts & then expand into parts. Your thoughts

 

[hj]

Is this all altuistic, or is there some kind of charge AND guarantee about your replies?

 

[Asktom]

It would be simpler to start with the stem and move back to the brand/model. There are a lot of faucets in this day and age that look the same and you will have people claiming the brand they have is UPC.

 

[satty]

Is this all altuistic, or is there some kind of charge AND guarantee about your replies?

No we plan to make money by identifying the exact make and model every time the user takes a picture of the faucet. We are @ about 95% accuracy.
-Satty

 

[satty]

It would be simpler to start with the stem and move back to the brand/model. There are a lot of faucets in this day and age that look the same and you will have people claiming the brand they have is UPC.

Thanks for the feedback Tom!! Appreciate it.
-Satty