1
00:00:00,000 --> 00:00:02,220
A supplier recalls one lot of peanut

2
00:00:02,220 --> 00:00:02,960
butter.

3
00:00:02,960 --> 00:00:04,860
You run a small food business.

4
00:00:04,860 --> 00:00:07,040
Your receiving spreadsheet, batch sheets, and shipment

5
00:00:07,040 --> 00:00:09,800
export each hold part of the answer.

6
00:00:09,800 --> 00:00:11,380
Which products did it reach?

7
00:00:11,380 --> 00:00:12,440
Who received them?

8
00:00:12,440 --> 00:00:14,677
And what are you still missing?

9
00:00:14,677 --> 00:00:18,577
This is RecallRoom, created by Shivam Gupta.

10
00:00:18,577 --> 00:00:21,357
It brings the evidence into one investigation.

11
00:00:21,357 --> 00:00:24,626
Every record in this demonstration is fictional.

12
00:00:24,626 --> 00:00:27,526
Start with the recalled ingredient.

13
00:00:27,526 --> 00:00:30,006
Follow its path through production to customers.

14
00:00:30,006 --> 00:00:33,606
480 oat bars and 320 energy bites.

15
00:00:33,606 --> 00:00:35,826
800 affected finished units.

16
00:00:35,826 --> 00:00:38,226
600 have shipped to three customers.

17
00:00:38,226 --> 00:00:40,551
200 remain on site.

18
00:00:40,551 --> 00:00:42,591
Click a shipment and you can inspect

19
00:00:42,591 --> 00:00:44,558
the source behind it.

20
00:00:44,558 --> 00:00:47,738
But, this cookie batch only says PB

21
00:00:47,738 --> 00:00:50,618
September That could refer to a different

22
00:00:50,618 --> 00:00:52,398
ingredient lot.

23
00:00:52,398 --> 00:00:55,198
RecallRoom keeps 240 cookies on a separate

24
00:00:55,198 --> 00:00:59,822
precautionary hold, including 120 that have shipped.

25
00:00:59,822 --> 00:01:03,262
Now open the supporting clarification.

26
00:01:03,262 --> 00:01:06,082
It identifies Lot B and explicitly says

27
00:01:06,082 --> 00:01:08,122
no Lot A was used.

28
00:01:08,122 --> 00:01:11,062
A reviewer excludes the Lot A relationship,

29
00:01:11,062 --> 00:01:14,866
then confirms Lot B using that evidence.

30
00:01:14,866 --> 00:01:18,806
The cookie hold leaves this recall path.

31
00:01:18,806 --> 00:01:21,985
The original relationship, exact quote, and review

32
00:01:21,985 --> 00:01:24,886
decision remain in the history.

33
00:01:24,886 --> 00:01:27,506
The recorded balance now shows 50 kilograms

34
00:01:27,506 --> 00:01:30,685
of the recalled ingredient remaining.

35
00:01:30,685 --> 00:01:33,565
Here is a real extraction with NVIDIA

36
00:01:33,565 --> 00:01:36,465
Nemotron on Nebius Token Factory.

37
00:01:36,465 --> 00:01:39,525
The model proposes records and exact source

38
00:01:39,525 --> 00:01:40,925
quotes.

39
00:01:40,925 --> 00:01:44,045
Application code validates the evidence and calculates

40
00:01:44,045 --> 00:01:45,365
the scope.

41
00:01:45,365 --> 00:01:48,845
A person reviews the proposal before import.

42
00:01:48,845 --> 00:01:51,845
The run history shows the actual provider,

43
00:01:51,845 --> 00:01:54,593
model, token usage, and latency.

44
00:01:54,593 --> 00:01:58,113
Finally, export the investigation packet.

45
00:01:58,113 --> 00:02:01,153
It includes the shipment register, evidence, review

46
00:02:01,153 --> 00:02:04,493
decisions, and customer specific message drafts.

47
00:02:04,493 --> 00:02:07,794
A quality manager controls what happens next.

48
00:02:07,794 --> 00:02:11,114
RecallRoom is designed for smaller teams using

49
00:02:11,114 --> 00:02:13,655
the records they already have.

50
00:02:13,655 --> 00:02:16,215
Recurring drills can expose missing evidence before

51
00:02:16,215 --> 00:02:17,834
an emergency.

52
00:02:17,834 --> 00:02:19,914
Our next step is to validate that

53
00:02:19,914 --> 00:02:22,781
workflow with quality professionals.

54
00:02:22,781 --> 00:02:25,921
A recall starts with one lot.

55
00:02:25,921 --> 00:02:27,601
RecallRoom helps you see where it ends

56
00:02:27,601 --> 00:02:30,225
and what still needs an answer.