Example projectWeb appAICustom LLM
Vehicle damage photo model
Hypothetical. Not a client, not a result.
// The problem
Every auto claim photo is opened by an adjuster to decide whether the file is a quick repair or a likely total loss, which slows the simple ones down.
// What we would build
A custom vision model fine-tuned on the insurer's own closed claims photos, served in its own AWS account, that sorts new files by likely severity and the parts affected so the adjuster decides faster. A general frontier model does not know your repair outcomes; yours does.
// What is in it
- A model fine-tuned on your past photos and how each claim was settled
- Likely severity and affected parts shown to the adjuster, never sent to the customer
- The adjuster decides; the model routes and suggests
- Measured against held-out claims before launch and checked for drift after
- Served in your own cloud account, with photos kept there
// Stack
- PyTorch
- Python
- Amazon S3
- Next.js
- AWS GPU instances
// estimate
- Build
- ≈ US$170,000 to US$259,000, delivered within 38 weeksCAD 241,800 to 369,200
- 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 3.9 weeks
- Core features
- 2.4 to 3.4 weeks
- Full build
- 1.4 to 2.4 weeks
- Training
- 9.9 to 16.9 weeks
- Testing and fixes
- 0.4 weeks
- Evaluation and red-teaming
- 3.9 to 4.4 weeks
- Launch
- 0.9 to 1.4 weeks
- Production
- 0.9 to 1.4 weeks
How it is paid
- Deposit 20%
- CAD 48,120 to 73,600
- Discovery 1.3%
- CAD 3,127.80 to 4,784.00
- Data and evaluation plan 8.2%
- CAD 19,729.20 to 30,176.00
- Design approved 1.3%
- CAD 3,127.80 to 4,784.00
- Core features 4%
- CAD 9,624 to 14,720
- Full build 2.6%
- CAD 6,255.60 to 9,568.00
- Training 24.8%
- CAD 59,668.80 to 91,264.00
- Testing and fixes 1.3%
- CAD 3,127.80 to 4,784.00
- Evaluation and red-teaming 16.5%
- CAD 39,699 to 60,720
- Launch 1.3%
- CAD 3,127.80 to 4,784.00
- Production 8.7%
- CAD 20,932.20 to 32,016.00
- Holdback, 30 days after launch (10%)
- CAD 24,060 to 36,800
- GPU time for training and testing, at cost
- CAD 1,200. Billed up front, at cost, outside the milestones