Skip to content
AtheronLABS

You're visiting from the United States. Prices are shown in US dollars. Not right?

labs@atheron:~/industries/waste-recycling/ai$ train agent --grounded

AI for recycling

Contamination spotted on truck cameras, then confirmed.

Contamination is a few addresses doing the wrong thing repeatedly. A camera and a model trained on your own loads can find those addresses, so education goes where it is needed. The model flags; a reviewer confirms before any notice goes out.

// what matters here

What is different in this industry.

01

Trained on your program

What counts as contamination depends on your program's rules. We fine-tune the model on your own labelled images, which is why it is custom.

02

Tied to the address

Each flag is joined to the cart's RFID lift record, so a notice goes to the household that tipped the cart.

03

A person confirms

A reviewer confirms flags before any notice is sent, and overrules become training data for the next round.

04

Measured before it is used

The model is scored on a held-out set of your images, and the agreed result is the gate for going live.

// example projects

Priced examples for this service.

Each one is hypothetical, labelled as such, and priced live from our rate card. Open any of them in the estimator and make it yours.

Example projectWeb appAICustom LLM

Contamination model on trucks

Hypothetical. Not a client, not a result.

// The problem

A recycling program loses loads to contamination, but cannot tell which routes or addresses are causing it, so education goes everywhere and changes nothing.

// What we would build

A custom vision model, fine-tuned on the program's own hopper camera images, that flags likely contamination as each cart is tipped, tied to the cart's RFID lift record, so the program can send targeted notices to the addresses that need them.

// What is in it

  • Hopper camera images captured at each lift
  • Likely contamination flagged by type, for a reviewer to check
  • Flags tied to the address through the cart's RFID lift record
  • Targeted notices and follow-up for repeat addresses
  • A model fine-tuned on your images and measured on a held-out set

// Stack

  • PyTorch
  • ONNX Runtime
  • Python
  • Next.js
  • AWS GPU instances

// estimate

Build
≈ US$168,000 to US$257,000, delivered within 41 weeksCAD 239,900 to 366,300
Hosting
≈ US$7,020 a monthCAD 10,000 a month
Support
≈ US$2,500 a monthCAD 3,565 a month

Prices in your currency are estimates from today's Bank of Canada rate. All invoicing is in CAD or USD.

AI route
Custom LLM
Model running cost
≈ US$6,530 a monthCAD 9,300 a month

Timeline by milestone

Discovery
2.9 to 3.4 weeks
Data and evaluation plan
0.4 weeks
Design approved
2.4 to 4.4 weeks
Core features
1.9 to 3.4 weeks
Full build
1.4 to 2.4 weeks
Training
8.9 to 15.4 weeks
Testing and fixes
0.4 weeks
Evaluation and red-teaming
3.9 to 4.4 weeks
Report checked against a past period
3.9 to 4.4 weeks
Launch
0.9 to 1.4 weeks
Production
0.9 to 1.4 weeks

Outside our hands, and added to the calendar

Truck readers and cart tags delivery
2 to 6 weeks
Fleet GPS API access from your provider
1 to 4 weeks

How it is paid

Deposit 20%
CAD 47,740 to 73,020
Discovery 1.1%
CAD 2,625.70 to 4,016.10
Data and evaluation plan 7.2%
CAD 17,186.40 to 26,287.20
Design approved 1.1%
CAD 2,625.70 to 4,016.10
Core features 3.5%
CAD 8,354.50 to 12,778.50
Full build 2.3%
CAD 5,490.10 to 8,397.30
Training 21.7%
CAD 51,797.90 to 79,226.70
Testing and fixes 1.1%
CAD 2,625.70 to 4,016.10
Evaluation and red-teaming 14.4%
CAD 34,372.80 to 52,574.40
Report checked against a past period 8.7%
CAD 20,766.90 to 31,763.70
Launch 1.1%
CAD 2,625.70 to 4,016.10
Production 7.8%
CAD 18,618.60 to 28,477.80
Holdback, 30 days after launch (10%)
CAD 23,870 to 36,510
GPU time for training and testing, at cost
CAD 1,200. Billed up front, at cost, outside the milestones

// questions

Questions we get about this.

Do contamination cameras record residents?

Cameras face into the hopper, not the street. Your privacy team reviews camera placement and retention before launch.

Does it need RFID carts to work?

To tie a flag to an address, yes, or very precise GPS. Without either, it still reports contamination by route.

Can it run on the truck?

Images can be processed on the truck or sent for processing in your cloud. We choose based on your trucks' connectivity and cost.

// next

Not quite your project?

Tell us what you have in mind. We will come back to you with a range and the questions that would narrow it. Or book a call and talk it through.

AI contamination detection for recycling collection programs | Atheron Network Labs