Sentinel
Home Dashboard Data Assistant
Built by NxtOps for Bharat Builds Tour

Find the money your business is quietly losing.

Sentinel reads your sales, stock, suppliers and payments. It then explains what is going wrong, why, and what to check first.

The dashboard says you're fine.

Small shops rarely lose money in one big mistake. They lose it in dozens of small ones that nobody has time to notice.

From raw records to root cause.

Four steps, no setup, and you stay in control the whole way.

1

Bring your data

Add records by hand or import CSVs: products, suppliers, sales, deliveries, returns and payments.

2

We investigate

Detectors scan for dead stock, stockouts, late suppliers, price creep, return spikes and slow payers.

3

We connect the dots

Related events are chained into a likely root cause, with the evidence behind every claim.

4

You decide

Every finding says what to check first. Nothing changes in your business unless you change it.

Six kinds of quiet loss.

Each one is small on its own. Together they decide whether the month was good.

Root-cause chains

A late supplier empties a shelf, the shelf costs you sales, and the sales show up as a vague dip. We link those events with dates and numbers, so you see the cause and not only the symptom.

Hidden stockouts

The days you could not sell, and what they cost you in gross profit.

Dead stock

Cash parked on shelves where nothing moves for months.

Price creep

Supplier costs that rose quietly while your selling price stayed put.

Return spikes

The one product or variant that keeps coming back.

Late payers

Customers who stretch your cash cycle a little more each time.

Designed to run on AWS.

A serverless pipeline with an agent at the centre. This file is the working prototype of the product layer.

UploadAmazon S3
IngestAWS Lambda
StateDynamoDB
RetrieveOpenSearch Serverless
InvestigateStrands agent, Bedrock
PresentCloudFront
Cognito sign-in and rolesEventBridge runs analysis when new data landsStep Functions orchestrates the pipelineCloudWatch monitoring and cost

Tools the agent can call

  • query_sales reads demand by product and day
  • supplier_scorecard compares promised and actual delivery
  • find_stockouts lists the days a product was unavailable
  • link_events connects causes to effects by date
  • search_history retrieves similar past cases
goal what is the biggest silent loss?
call supplier_scorecard()
  Apex: promised 3d, actual 5.2d
call find_stockouts()
  6 stockouts, 30 days out
call link_events(stockouts, deliveries)
  6 of 6 follow a late delivery
answer late deliveries likely caused the
  stockouts. Review reorder points first.
In this file today
In production on AWS
Detectors run in your browser
Lambda and Step Functions
Records live in local browser storage
DynamoDB for state, S3 for uploads
Search runs in memory
OpenSearch Serverless for retrieval
Assistant calls the model with your own key
Strands agent on Bedrock, key stays server-side
No sign-in
Cognito with per-shop access

The demo shop's numbers are synthetic. Your own data never leaves your browser unless you connect the assistant to a model provider.

See it find the leaks in seconds.

Try the demo shop, or replace it with your own numbers.

Sentinel by NxtOps for Bharat Builds Tour. The demo shop uses synthetic data.

Silent lossesDemo data

Findings

Stockout timeline

Weekly units sold, with the days you ran out and when deliveries landed.

Units sold per weekOut of stockOn-time deliveryLate delivery

Supplier scorecard

Promised delivery time against what actually happened.

Your data

Add records one at a time or import a CSV. The analysis updates as soon as you save.

Click any row to edit it.

Ask anything.

Grounded in your data. Curious about something else? Ask that too.