Numi v3: Automated AI Coin Sorting Prototype

AlbumNerd

Last summer, I shared my journey of building an AI-powered coin sorting robot, Numi v2. After several months of development, I paused the project because AI technology wasn't quite ready. However, by March, AI advancements convinced me it was time to pick it back up.

The past two months have been an intense sprint to complete Numi v3, just in time for the Colorado Springs Coin Club Show and the ANA Summer Seminar. I'm thrilled to announce that after hundreds of hours of work, Numi v3 is ready for its public debut!

I'll be demoing Numi at:

• Colorado Springs Coin Club Show [June 27th–28th]
• ANA Summer Seminar [June 28th–July 3rd]

If you can’t attend, check out this video demo!

Why I Created Numi
I'm a passionate coin collector and technologist [former technology product manager]. I believe strongly in the potential of Artificial Intelligence to positively impact our hobby. Many people still underestimate the potential benefits of AI to the hobby.

Numi v3 will initially focus on Lincoln Wheat cents, sorting them efficiently by year and mint mark.

How Numi Works
Numi v3's sorting chassis is constructed from over 500 LEGO pieces, powered internally by a Raspberry Pi 5. I wrote the software in Python with help from Cursor AI, totaling approximately 14,000 lines of code.

When starting Numi, you'll first encounter a welcome page offering two operational modes: Standard Mode and Dealer Mode.

Standard Mode handles Numi’s essential functions:

• Controlling LEGO motors to guide coins smoothly through the machine
• Using dual cameras to capture high-resolution photos when coins enter the viewing area
• Basic user interface for displaying and saving photos securely to Google Cloud
• AI analysis of coin images via Google Gemini 2.5 Flash (a powerful and affordable image recognition AI from Google)
• Automated sorting by decade and year
• Fully automated looping process for continuous sorting

Creating Standard Mode was by far the biggest challenge. I had to meticulously engineer every mechanical detail and software component from scratch. Once established, it laid the groundwork for easily adding other useful modes.

Dealer Mode, designed in collaboration with feedback from coin dealers nationwide, enables dealers to quickly identify valuable coins within bulk lots.

In Dealer Mode, you start by specifying coins of interest through a custom watch list:

Once running, Numi alerts you when it identifies coins matching your list, capturing detailed images and tracking analysis costs (approximately $6 per 10,000 coins analyzed). Numi can analyze roughly one coin every 15 seconds, or around 5,000 coins per day.

After processing, you can generate a custom Google Sheet report listing discovered coins along with image links.

Dealer Mode is configured for real-time feedback at demonstrations, helping me refine Numi based on practical insights from actual dealers.

With the core system in place, developing additional modes is straightforward. One idea I'm particularly excited about is Collector Mode, which would help collectors complete coin albums interactively. Users would see a digital album highlighting missing coins, dynamically updated as Numi sorts and identifies them.

I’m eager to hear your ideas or additional scenarios you'd find helpful!

Consider Numi v3 a working prototype and proof of concept. My main goal at upcoming events is to gather community feedback on the broader impact of automated AI coin sorting and its future potential.

This project has undergone significant evolution over the past two years. Initially, Numi v1 was a customized ChatGPT-powered coin grader. Numi v2 automated silver coin sorting from circulation. Each iteration brought valuable lessons that shaped this latest version. The goal has already remained the same though: to make AI genuinely beneficial and approachable for everyone in our hobby.

Looking ahead, I'll focus on refining Numi based on real-world tests. Collectors from around the country have reached out with interest in helping them sort through their bulk wheat hoards. I’ve had dealers offer to have me come in and run through their bulk wheaties to get concrete numbers on how time is saved and extra revenue generated.

Thank you for following my journey with Numi v3. I hope to meet you at the upcoming show or seminar!

Justin, aka Dansco Dude

braddick

Excellent work Justin.
I wish you nothing but unbridled success.

yosclimber

It's cool you have been having fun doing computer vision and Lego stuff.

I still do not see it as a product that collectors or dealers would want to use.
This is because of the nature of "bulk wheat hoards".
It is closely related to the "unsearched wheat rolls" that are often advertised.
I don't think there are many 1909-S or 1914-D coins in those hoards.
When those were constructed, somebody looked at the coins, so they cannot be truly unsearched.
They looked enough to at least say they were Wheat cents and not Memorial cents or Indian cents.
Most of the time I believe the person would also check the date to see if it was valuable.

Still, it's a fun challenge to see if you can make a machine do something that many collectors enjoy doing.

TomB

Pretty dang cool!

jmlanzaf

@yosclimber said:
It's cool you have been having fun doing computer vision and Lego stuff.

I still do not see it as a product that collectors or dealers would want to use.
This is because of the nature of "bulk wheat hoards".
It is closely related to the "unsearched wheat rolls" that are often advertised.
I don't think there are many 1909-S or 1914-D coins in those hoards.
When those were constructed, somebody looked at the coins, so they cannot be truly unsearched.
They looked enough to at least say they were Wheat cents and not Memorial cents or Indian cents.
Most of the time I believe the person would also check the date to see if it was valuable.

Still, it's a fun challenge to see if you can make a machine do something that many collectors enjoy doing.

There are some dealers that still sort coin hoards. There are definitely a lot of collectors who do.

jesbroken

Interesting. I wish you the best. As far as hoards or roll hunting goes, I agree with @yosclimber as to rarity of coins in these "unsearched" group of coins.
Now if perfected to find varieties and attributions, while sad, might be a useable avenue to go down.
Regardless, good luck.
Jim

jmlanzaf

@jesbroken said:
Interesting. I wish you the best. As far as hoards or roll hunting goes, I agree with @yosclimber as to rarity of coins in these "unsearched" group of coins.
Now if perfected to find varieties and attributions, while sad, might be a useable avenue to go down.
Regardless, good luck.
Jim

You don't have to directly find the variety to find the years that merit closer look.

Mr_Spud

I think it’s great you are using legos for this purpose. I’m picturing a coin dealer offering a “free” version of a Coinstar machine that doesn’t charge people to cash in their change with one of your Numismatic sorters built inside it and the dealer gets to keep the ones worth a premium. Or the dealer can charge extra for people to use the CoinstarNumi machine and the Numi catches anything worth more than face value and gives those ones back with a printout saying what they are.

GaCoinGuy

Neat concept.

Are the coins loaded in by hand one-at-a-time or will it have a hopper of some sort to load up and self-feed?

Brings to mind a business partner of my mom's that was, with her brother's and sisters, cleaning out the family home (19th century built) and they had accumulated a few 5-gallon buckets of loose coins with the oldest coin I heard of being an 1834 half cent coin. Would be nice to have something that could scan and catalogue something along that line without causing any further damage to the coins.

cmerlo1

That is awesome- let us know when/if you ever bring it to market.

I've been brought in to help dealer friends of mine sort bulk Lincon cent and silver purchases, including searching for varieties and this would be a giant help in at least separating the years with good varieties, for example Lincoln cents years 1909, 1910-S, 1911-D, 1911-S, 1917, 1925-S, 1934, 1934-D, 1943, 1943-D, 1943-S, 1944-D, 1947, 1955, etc.

pcgsregistrycollector

Awesome new program!

Namvet69

Impressive demo and proof of concept. Best of luck as you continue to develop your system. Peace Roy

AlbumNerd

@yosclimber said:
It's cool you have been having fun doing computer vision and Lego stuff.

I still do not see it as a product that collectors or dealers would want to use.
This is because of the nature of "bulk wheat hoards".
It is closely related to the "unsearched wheat rolls" that are often advertised.
I don't think there are many 1909-S or 1914-D coins in those hoards.
When those were constructed, somebody looked at the coins, so they cannot be truly unsearched.
They looked enough to at least say they were Wheat cents and not Memorial cents or Indian cents.
Most of the time I believe the person would also check the date to see if it was valuable.

Still, it's a fun challenge to see if you can make a machine do something that many collectors enjoy doing.

I can see where you're coming from yosclimber. The idea is to start with a narrow scope to a use case that I know some coin dealers and wheat collectors are currently dealing with. I'm generally pessimistic about actually finding key and semi-key dates in circulation, but I want first to obtain concrete numbers and results to validate this. I think the flip side of finding varieties & errors is more promising. Numi isn't capable of detecting variety and errors itself, but it can sort by year and mint mark.

I'm speaking with a wheat collector who has a million wheaties that Numi could sort through. He wants to look for the 1984 double die. Numi would separate all 1984 cents [along with other possible years with varieties/errors] and capture photos. The collector could examine the high-resolution photos to see which piles are worth further examination. It would be an interesting use case where I could walk away answering:

• How much does it cost to sort 1 million coins
• How much time did it take to sort 1 million coins
• How much the collector made, if anything at all

I also have a few coin shops who have offered to let me drop by and sort through their wheaties. I would pay for the expenses of travel and sorting. Getting the data is the most valuable part.

@Mr_Spud said:
I think it’s great you are using legos for this purpose. I’m picturing a coin dealer offering a “free” version of a Coinstar machine that doesn’t charge people to cash in their change with one of your Numismatic sorters built inside it and the dealer gets to keep the ones worth a premium. Or the dealer can charge extra for people to use the CoinstarNumi machine and the Numi catches anything worth more than face value and gives those ones back with a printout saying what they are.

This is a really interesting idea. I bet some dealors would entertain the idea of pointing a customer to a machine and say "Just put your coins through there. I'll let you know if you have anything I'm interested in."

@GaCoinGuy said:
Neat concept.

Are the coins loaded in by hand one-at-a-time or will it have a hopper of some sort to load up and self-feed?

Numi uses a self-feed system. Currently you can drop a max of 30 coins without it jamming [usually]. My hope is to drastically boost that number.

PeakRarities

@AlbumNerd said:

I also have a few coin shops who have offered to let me drop by and sort through their wheaties. I would pay for the expenses of travel and sorting. Getting the data is the most valuable part.

My, how generous of them to let you do that 🫤.

I think I'd at least offer a majority portion of the equivalent wages for the time, if I had enough wheat cents to be worth the effort. For example, 20 hours for a $20/hr employee would cost $200, I'd hope they'd offer at least 50% of that considering your expenses. That, or maybe a cut of the profit from the sale of any rare varieties found.

AlbumNerd

@PeakRarities said:

@AlbumNerd said:

I also have a few coin shops who have offered to let me drop by and sort through their wheaties. I would pay for the expenses of travel and sorting. Getting the data is the most valuable part.

My, how generous of them to let you do that 🫤.

I think I'd at least offer a majority portion of the equivalent wages for the time, if I had enough wheat cents to be worth the effort. For example, 20 hours for a $20/hr employee would cost $200, I'd hope they'd offer at least 50% of that considering your expenses. That, or maybe a cut of the profit from the sale of any rare varieties found.

The way I think about it, the data I walk away with would be more valuable than any money I would get through profit sharing at this early stage in prototyping. The more money I find for them, the better it is for Numi's marketing. I get to go around and say, "Numi helped Norman's coin shop in Albany sort 30,000 wheat cents and extract an extra $3,000 in inventory within a week. It only costs $20 to run the batch. You can reach out to him and ask him what he thinks about Numi"

The data, feedback, and possible testimonials are more than I would ever get back in profit sharing at this point.

Except if they find a 1958/7 double die worth $250,000. Then I would ask them to cover the AI analysis costs

Dave99B

Well done! I watched your demo. Pretty exciting technology!!!

Dave

PeakRarities

@AlbumNerd said:

@PeakRarities said:

@AlbumNerd said:

I also have a few coin shops who have offered to let me drop by and sort through their wheaties. I would pay for the expenses of travel and sorting. Getting the data is the most valuable part.

My, how generous of them to let you do that 🫤.

I think I'd at least offer a majority portion of the equivalent wages for the time, if I had enough wheat cents to be worth the effort. For example, 20 hours for a $20/hr employee would cost $200, I'd hope they'd offer at least 50% of that considering your expenses. That, or maybe a cut of the profit from the sale of any rare varieties found.

The way I think about it, the data I walk away with would be more valuable than any money I would get through profit sharing at this early stage in prototyping. The more money I find for them, the better it is for Numi's marketing. I get to go around and say, "Numi helped Norman's coin shop in Albany sort 30,000 wheat cents and extract an extra $3,000 in inventory within a week. It only costs $20 to run the batch. You can reach out to him and ask him what he thinks about Numi"

The data, feedback, and possible testimonials are more than I would ever get back in profit sharing at this point.

Except if they find a 1958/7 double die worth $250,000. Then I would ask them to cover the AI analysis costs

That sounds a bit better, I don't think the data alone would make it worthwhile, but as long as they'd be comfortable with you marketing their testimonials I can see that having substantial marketing value. Good luck, keep us posted!

davewesen

Very nice - a computerized 2020's version of the scan-o-matic from the 50's-60's.

If you improve it enough be able sort and grade BU rolls, it will be very marketable.