Granted Patent B2

US 12,406,295 · App. 18/509,157 · Granted Sep 2, 2025

Inventors: Mingzhe Zhuang (Los Altos Hills, CA); Camille van Horne (San Francisco, CA); Christopher Rudnick (Pacifica, CA); Benjamin Knight (Oakland, CA); Chris Jenkins (Thornton, CO); Viktoriya Andonova (San Francisco, CA); Djordje Gluhovic (Oakland, CA); Riddhima Sejpal (Dublin, CA); Maksim Golivkin (Oakland, CA); Sharath Rao Karikurve (Berkeley, CA)

Assignee: Maplebear Inc.

G06Q30/0635G06Q30/0639

Abstract

Claims (106)

1. A computer-implemented method performed by one or more processors of an online system, the computer-implemented method comprising:

training a machine-learned availability model for determining availabilities of items in warehouses, wherein training of the machine-learned availability model comprises:

applying a plurality of training samples to train the machine-learned availability model, the plurality of training samples comprising previous delivery orders, wherein at least one training sample comprises a training label indicating whether an item was found at a warehouse and a plurality of characteristics associated with the item;

receiving a first set of notifications regarding availabilities of a plurality of items, wherein a notification includes whether one or more items were found at the warehouse;

determining, based on the first set of notifications, that a set of items were reported to be not found, the set of items being reported to be not found for at least a threshold number of times;

retraining, based on the set of items, the machine-learned availability model to improve accuracy of predictions of the machine-learned availability model, wherein retraining the machine-learned availability model comprises:

generating an additional set of training samples using the set of items that were not found for at least the threshold number of times,

inputting the additional training samples to the machine-learned availability model for the machine-learned availability model to generate a confidence score indicative of an accuracy of a prediction of availability of one or more items in the set,

responsive to the confidence score indicative of the accuracy of the prediction of availability of the specific item being below a threshold, sending instructions to indicate the availability of the one or more items need to be verified,

receiving a second set of notifications indicating one or more items in the set of items were found in the warehouse,

updating the additional set of training samples based on the second set of notifications,

applying the additional set of training samples to the machine-learned availability model, and

adjusting weights of the machine-learned availability model based on the applying of the additional set of training samples;

receiving, from a user device, a selection of a particular item;

applying the retrained machine-learned availability model to generate a prediction of whether the particular item is available at the warehouse; and

causing to display a result based on the prediction generated by the machine-learned availability model regarding the particular item.

2. The computer-implemented method of claim 1 , wherein receiving the first set of notifications regarding availabilities of the plurality of items comprises:

receiving an indication that a shopper has obtained at least one of the plurality of items;

determining that the obtained one of the plurality of items is a difficult-to-find item that was not found for at least the threshold number of times;

updating the training samples for the machine-learned availability model to indicate that the found one of the plurality of items is a difficult-to-find item.

3. The computer-implemented method of claim 1 , wherein receiving the first set of notifications regarding availabilities of the plurality of items comprises:

receiving an indication that a shopper could not find one of the plurality of items;

designating the one of the plurality of items as an unfound item;

determining that the unfound item is a difficult-to-find item that was not found for at least the threshold number of times;

updating the additional training samples for the unfound item for the machine-learned availability model to indicate that the found one of the plurality of items is a difficult-to-find item.

4. The computer-implemented method of claim 1 , wherein causing to display a result based on the prediction generated by the machine-learned availability model regarding the particular item comprises:

prompting a shopper to provide information for finding the particular item in the warehouse;

receiving a response from the shopper whether the shopper is able to find the particular item in the warehouse; and

causing the user device to display an availability of the particular item in the warehouse.

5. The computer-implemented method of claim 4 , wherein the information for finding the particular item in the warehouse comprises one or more of: a picture of a location in the warehouse of the particular item, a text description of a location in the warehouse of the particular item, or any combination thereof.

6. The computer-implemented method of claim 1 , wherein applying the machine-learned availability model to generate a prediction of whether the particular item is available comprises:

generating a predicted availability score associated with the particular item;

comparing the predicted availability score to an availability threshold; and

determining the prediction of whether the particular item is available based on the comparison.

7. The computer-implemented method of claim 1 , further comprising:

receiving information from a first shopper that the first shopper is able to find the particular item at the warehouse;

receiving an additional order that includes the particular item; and

transmitting the information to a second shopper for finding the particular item at the warehouse.

8. The computer-implemented method of claim 7 , wherein receiving information from the first shopper comprises receiving a photo taken by the first shopper locating the particular item.

9. The computer-implemented method of claim 1 , further comprising:

receiving a threshold number of negative indications that an item is not found at the warehouse; and

responsive to receiving the threshold number of negative indications, removing the item from an offering list of the online system with respect to the warehouse.

10. A non-transitory computer-readable medium configured to store code comprising instructions, wherein the instructions, when executed by one or more processors, cause the one or more processors to:

train a machine-learned availability model for determining availabilities of items in warehouses, wherein training of the machine-learned availability model comprises:

applying a plurality of training samples to train the machine-learned availability model, the plurality of training samples comprising previous delivery orders, wherein at least one training sample comprises a training label indicating whether an item was found at a warehouse and a plurality of characteristics associated with the item;

receive a first set of notifications regarding availabilities of a plurality of items, wherein a notification includes whether one or more items were found at the warehouse;

determine, based on the first set of notifications, that a set of items were reported to be not found, the set of items being reported to be not found for at least a threshold number of times;

retrain, based on the set of items, the machine-learned availability model to improve accuracy of predictions of the machine-learned availability model, wherein retraining the machine-learned availability model comprises:

generating an additional set of training samples using the set of items that were not found for at least the threshold number of times,

inputting the additional training samples to the machine-learned availability model for the machine-learned availability model to generate a confidence score indicative of an accuracy of a prediction of availability of one or more items in the set,

responsive to the confidence score indicative of the accuracy of the prediction of availability of the specific item being below a threshold, sending instructions to indicate the availability of the one or more items need to be verified,

receiving a second set of notifications indicating one or more items in the set of items were found in the warehouse,

updating the additional set of training samples based on the second set of notifications,

applying the additional set of training samples to the machine-learned availability model, and

adjusting weights of the machine-learned availability model based on the applying of the additional set of training samples;

receive, from a user device, a selection of a particular item;

apply the retrained machine-learned availability model to generate a prediction of whether the particular item is available at the warehouse; and

cause to display a result based on the prediction generated by the machine-learned availability model regarding the particular item.

11. The non-transitory computer-readable medium of claim 10 , wherein the instruction to receive the first set of notifications regarding availabilities of the plurality of items comprises instructions to:

receive an indication that a shopper has obtained at least one of the plurality of items;

determine that the obtained one of the plurality of items is a difficult-to-find item that was not found for at least the threshold number of times;

update the training samples for the machine-learned availability model to indicate that the found one of the plurality of items is a difficult-to-find item.

12. The non-transitory computer-readable medium of claim 10 , wherein the instruction to receive the first set of notifications regarding availabilities of the plurality of items comprises instructions to:

receive an indication that a shopper could not find one of the plurality of items;

designate the one of the plurality of items as an unfound item;

determine that the unfound item is a difficult-to-find item that was not found for at least the threshold number of times;

update the additional training samples for the unfound item for the machine-learned availability model to indicate that the found one of the plurality of items is a difficult-to-find item.

13. The non-transitory computer-readable medium of claim 10 , wherein the instruction to cause to display a result based on the prediction generated by the machine-learned availability model regarding the particular item comprises instructions to:

prompt a shopper to provide information for finding the particular item in the warehouse;

receive a response from the shopper whether the shopper is able to find the particular item in the warehouse; and

cause the user device to display an availability of the particular item in the warehouse.

14. The non-transitory computer-readable medium of claim 13 , wherein the information for finding the particular item in the warehouse comprises one or more of: a picture of a location in the warehouse of the particular item, a text description of a location in the warehouse of the particular item, or any combination thereof.

15. The non-transitory computer-readable medium of claim 10 , wherein the instruction to applying the machine-learned availability model to generate a prediction of whether the particular item is available comprises instructions to:

generate a predicted availability score associated with the particular item;

compare the predicted availability score to an availability threshold; and

determine the prediction of whether the particular item is available based on the comparison.

16. The non-transitory computer-readable medium of claim 10 , wherein the instructions, when executed, further cause the one or more processors to:

receive information from a first shopper that the first shopper is able to find the particular item at the warehouse;

receive an additional order that includes the particular item; and

transmit the information to a second shopper for finding the particular item at the warehouse.

17. The non-transitory computer-readable medium of claim 16 , wherein the instruction to receive information from the first shopper comprises instructions to receive a photo taken by the first shopper locating the particular item.

18. The non-transitory computer-readable medium of claim 10 , wherein the instructions, when executed, further cause the one or more processors to:

receive a threshold number of negative indications that an item is not found at the warehouse; and

responsive to receiving the threshold number of negative indications, remove the item from an offering list of an online system with respect to the warehouse.

19. A system comprising:

one or more processors; and

memory configured to store code comprising instructions, wherein the instructions, when executed by one or more processors, cause the one or more processors to:

train a machine-learned availability model for determining availabilities of items in warehouses, wherein training of the machine-learned availability model comprises:

applying a plurality of training samples to train the machine-learned availability model, the plurality of training samples comprising previous delivery orders, wherein at least one training sample comprises a training label indicating whether an item was found at a warehouse and a plurality of characteristics associated with the item;

receive a first set of notifications regarding availabilities of a plurality of items, wherein a notification includes whether one or more items were found at the warehouse;

determine, based on the first set of notifications, that a set of items were reported to be not found, the set of items being reported to be not found for at least a threshold number of times;

retrain, based on the set of items, the machine-learned availability model to improve accuracy of predictions of the machine-learned availability model, wherein retraining the machine-learned availability model comprises:

generating an additional set of training samples using the set of items that were not found for at least the threshold number of times,

inputting the additional training samples to the machine-learned availability model for the machine-learned availability model to generate a confidence score indicative of an accuracy of a prediction of availability of one or more items in the set,

responsive to the confidence score indicative of the accuracy of the prediction of availability of the specific item being below a threshold, sending instructions to indicate the availability of the one or more items need to be verified,

receiving a second set of notifications indicating one or more items in the set of items were found in the warehouse,

updating the additional set of training samples based on the second set of notifications,

applying the additional set of training samples to the machine-learned availability model, and

adjusting weights of the machine-learned availability model based on the applying of the additional set of training samples;

receive, from a user device, a selection of a particular item;

apply the retrained machine-learned availability model to generate a prediction of whether the particular item is available at the warehouse; and

cause to display a result based on the prediction generated by the machine-learned availability model regarding the particular item.

20. The system of claim 19 , wherein the instruction to receive the first set of notifications regarding availabilities of the plurality of items comprises instructions to:

receive an indication that a shopper has obtained at least one of the plurality of items;

determine that the obtained one of the plurality of items is a difficult-to-find item that was not found for at least the threshold number of times;

update the training samples for the machine-learned availability model to indicate that the found one of the plurality of items is a difficult-to-find item.

Assignments (1)

Continuity (3)

References Cited (30)