Why can AI lower vision screening costs by 90%——The confession of AI-Optic's founder

Hi,my name is Wayne,the founder of AI-Optic. Here I want to tell you a true story. The story is about how an idea became a real product. I also want to tell you my plan for the future.

What I'm about to say is quite unique. You won't be able to find anything similar elsewhere. Just as the sub headline suggests,some contents are tech-related,but I'll try my best to make them simple and easy,so they won't slow you down.

Purpose

There're two purposes for this article. The first is to share our experiences and ideas. They might be helpful for your work. The second is a bit selfish. I think the more you know us,the more confident you will be to work with us. Moreover,in the end I'll offer an investment opportunity. Even if you don't have money to invest,you may still benefit from it.

Reading the article takes about 15 min. It's well worth your time. Trust me.

Background

The story started in 2018. I was looking for business opportunities when I discovered vision screener. Compared with regular autorefractor,vision screener has great advantages. It is portable,and the user doesn't need to be a licensed optician.

But the shortcoming?is also obvious,vision screener is very expensive. Take Spot vision screener?for example. In China it takes up to 15,000 USD. In the United States the price is lower,but still takes 7,800 USD. The high price seriously?limits the business potential,so I tried to find out why vision screener?is so expensive.

Reason

I figured out the reason after some time. In order to make it clear for?you,I need to explain vision screener's?working principle?first. The complete details?could bore you to death,so I give you the most concise version: under eccentric infrared light,if refractive errors are different,the shapes of the pupil are different. In other words,refractive error can be calculated from the shape of the pupil.

If you can't understand the principle completely,that's ok. You can still understand the?reason:inside vision screener,the calculations use traditional image-processing algorithms,which require industrial camera and high-performance CPU. Those two types of hardware are very expensive.

Seed

After figuring out the reason,I had no idea what to do next,so the vision screener thing stopped. One day in the summer of 2019,it came to me if the calculation uses AI model,vision screener can use cheap consumer camera, and if the AI model is deployed on a cloud server,high-performance CPU won't be necessary. This will lower the costs dramatically. And the good news didn't stop there. Because AI model is far more accurate,testing accuracy will increase a lot.

Over the next few months,this idea just grew bigger and stronger. I often wondered why it didn't come to me sooner. At the end of the year,I made up my mind to invest in it,to turn the idea into a real product. I gave the upcoming product a name---AI-Optic.

Hardware

Before jumping?into the details?of technical stuff,I needed to think carefully?about the selling price. This reminded me of Peter Thiel's book---Zero to One. I like the book very much and have read it several times. Inspired by Peter Thiel's theory,I decided the selling price must be under 1,000 USD.

The price included reseller's?profit,so I had to keep the manufacturing costs under 100 USD. This meant I couldn't use off-the-shelf components, and must design everything from scratch, including lens,camera,control board,driver software,infrared light board,etc.

To develop those components,there was a huge?amount of work to do,and the risk of failure was pretty high,so I hired three teams. Two of them failed,but one succeeded,and the manufacturing costs was only 65 USD. Then I hired another team to improve the design. In Jan.2022,we carried out small?scale production?successfully.

Data

The second job was to collect data for AI model training. In order to start the work sooner,we designed custom hardware?for data collection. Because a small number of machines should be enough,we didn't pay much attention to costs. Therefore we used off-the-shelf components,and finished the design in two months. In the end we assembled seven machines.

Collecting data is both trivial and troublesome,and we were prepared for this,but COVID-19 made the situation even worse. We had to pay nearly 50% more than expected. Data collection?started in Sep.2020. Over the next four months,we successfully collected 7,000 people's data,and I tested AI model's accuracy in the process. At the end of the year,I found the accuracy had met our requirement,so the second job was done.

AI model

Now you've come to a mysterious?place---AI model,which is the core of the project. Training AI model may sound very difficult,but actually it is way easier than most people think,thanks to open source software?in AI industry. The training work had already started during data collection,and I continued to improve the model after all the data were ready. In Mar.2021,I got the final version of our AI model.

I did the training work myself. My graduate major?was computer science,and since the summer of 2019,I'd learned?all the?latest stuffs in AI tech. so I could do the job well. In fact,I might be the best candidate?for this job,because I'm also familiar with vision screener,I know optometry well, and I designed the data format,all of which were very helpful for the training work.

Actually training AI model is a very pleasant?experience. The 7,000 people's data contain 630,000 images,take 1.7 TB in size. Training the model once takes 2-3 days. It feels like the best time of my life?to watch the accuracy jump on the screen.

Market

In Jan.2022,we finished all the development work,and got everything ready,so we carried out small?scale production. Then we tried online marketing?and got a small number of customers. These customers helped find some problems with our product. For example,the front cap may fall off during shipping;some Wi-Fi signals aren't compatible;the machine doesn't support printer,which could be inconvenient?sometimes.

We responded to these problems one by one:we changed the glue for the front cap,it won't fall off anymore;we studied Wi-Fi carefully,and worked out troubleshooting methods;we developed printing module,now the module is under testing.

As we keep fixing bugs in the product,AI-Optic becomes more and more mature.

Future

Now let's change the topic and talk about my plan for the future. There're two important jobs according to my plan.

The first is of course marketing and sales,this is the most important work for the next several months.

The second is new product development. After carrying out the first job successfully,I plan to develop two more AI products for eye care. Since AI can make vision screener much better,it can do the same for other eye care products.

New products

The first product I want to improve is fundus camera. After using AI tech,the new product can screen diabetic retinopathy?all by itself,there will be no need for eye doctor's help. The product is useful in many places,such as small hospital, nursing home,health examination center. Family doctor and diabetic?patients also can use it. Just like AI-Optic,I plan to bring the price down?by a big margin. The manufacturing costs will be kept under 200 USD,so the selling price can be under 1,500 USD.

The second product I want to improve?is retinal OCT. The new product can screen multiple eye diseases?automatically,eye doctor's participation won't be necessary. It can also be used as a regular OCT,but the price is way lower. I plan to keep the manufacturing costs under 2,000 USD,and the selling price under 10,000 USD.

If you don't know eye care well,the above contents may sound boring,but if you do,you should've?realized they mean a lot for eye care. I've spent a lot of time studying the two products,and every thought is based on solid research. I assure you?they are not daydreams.

Meaning

Now you've read almost everything. Since you've come this far,you'll probably agree AI-Optic is very promising. If Peter Thiel's book is right, in a few years AI-Optic will obtain?a?monopoly?position in the market. This means vision screening service will become universal. A?huge number of?students will benefit from it,especially the poor and the vulnerable.

If the two new products can be developed successfully,serious eye diseases can be found easily at an early stage. Patients can avoid unnecessary pain,save a lot of medical expenses,and most importantly,it can help prevent blindness from eye diseases. The?elderly?will benefit the most.

Investment

To carry out the plan?mentioned above,we need 3 million USD investments. In return,investors will hold?20%?of?the company's?share.

I write this article partly because I want to find potential investors,but first?I must say?investment has risk. If there's misjudgment?in what I've written,you and I may suffer financial loss. As you should've guessed,I invested most of my money in AI-Optic. In some extreme cases,the financial loss may be severe.

If my judgment is right,you and I will have a lot of financial returns. But unlike most investments,our society will benefit the most,far greater than you and me,because eye care will become radically more affordable. I am sure this can make you feel great,very great.

I might be right,or I might be wrong,either way,you need to make your own judgment.

Contact

If you are interested to invest,or if you have any questions,please do not hesitate to contact me. My email address is [email protected].

If your money is less than 3 million USD,your investment is welcome just the same. We'll continue to look for the rest of the money. But we won't be able to accept investment less than 100,000 USD,because private company in China can't have more than 200 shareholders. I'm sorry.

I understand most people don't have a large amount of money?to invest. But if you want to help and know someone who can invest in AI-Optic,please forward the message. If we receive the investment,we'll pay you 2% commission. Just contact?me first,so I can determine you're related to the investment.

If you don't know any potential investor but still want to help,please spread this message. I appreciate the effort and will always be grateful for your support.

If you have any suggestions for?our work,please do not hesitate to let me know. We're now in uncharted territory. Your advice could be invaluable.

I look forward to hearing from you soon.

End

Lastly,I'd like to say this:I believe our society will reward those who contribute to it. This was my leap of faith when I invested in AI-Optic.