R&D formulation assistant
Hypothetical. Not a client, not a result.
// The problem
A plastics compounder has decades of formulation experiments in notebooks and spreadsheets, and its chemists start each customer request by asking the one person who remembers what was tried.
// What we would build
A custom model fine-tuned on the company's own formulation records and lab results, which suggests starting formulations for a target property and finds similar past experiments, with the records behind each suggestion. It runs on a GPU server the company owns, inside its own network, because the formulations are the business. Chemists decide what to test.
// What is in it
- Formulation records and lab results cleaned and joined into one training set
- A model fine-tuned on your own data, run on a GPU server you own
- Starting formulations suggested for a target property, with the past records behind them
- Search across every past experiment by ingredient, property or result
- An evaluation set built with your chemists, run before every new version
// Stack
- Llama (fine-tuned)
- PyTorch
- vLLM
- PostgreSQL with pgvector
- Next.js
- Lenovo GPU server
// estimate
- Build
- ≈ US$159,000 to US$242,000, delivered within 34 weeksCAD 226,300 to 344,300
- Hosting
- ≈ US$176 a monthCAD 250 a month
- Support
- ≈ US$2,500 a monthCAD 3,565 a month
- Hardware
- ≈ US$47,800 to US$197,000CAD 68,000 to 280,800
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$176 a monthCAD 250 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 1.9 weeks
- Training
- 7.9 to 13.9 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
Outside our hands, and added to the calendar
- Hardware delivery, after it is ordered
- 2 to 6 weeks
How it is paid
- Deposit 20%
- CAD 45,020 to 68,620
- Discovery 1.3%
- CAD 2,926.30 to 4,460.30
- Data and evaluation plan 8.2%
- CAD 18,458.20 to 28,134.20
- Design approved 1.3%
- CAD 2,926.30 to 4,460.30
- Core features 4.1%
- CAD 9,229.10 to 14,067.10
- Full build 2.7%
- CAD 6,077.70 to 9,263.70
- Training 24.7%
- CAD 55,599.70 to 84,745.70
- Testing and fixes 1.3%
- CAD 2,926.30 to 4,460.30
- Evaluation and red-teaming 16.4%
- CAD 36,916.40 to 56,268.40
- Launch 1.3%
- CAD 2,926.30 to 4,460.30
- Production 8.7%
- CAD 19,583.70 to 29,849.70
- Holdback, 30 days after launch (10%)
- CAD 22,510 to 34,310
- GPU time for training and testing, at cost
- CAD 1,200. Billed up front, at cost, outside the milestones
- GPU server for AI, for the model
- CAD 68,000 to 280,800. Billed up front, at cost, outside the milestones