# SR&ED for AI & Machine Learning Companies | GrantOps

AI and machine learning development is among the most naturally SR&ED-eligible work in Canada. The experimental, iterative nature of ML research — training models with uncertain outcomes, developing novel architectures, and pushing the boundaries of what automated systems can do — aligns directly with CRA's definition of systematic investigation to resolve technological uncertainty.

## What AI/ML R&D Qualifies for SR&ED?

AI and ML projects that typically qualify include:

- Developing novel neural network architectures or training methodologies where performance outcomes are uncertain
- Building custom NLP models, computer vision systems, or recommendation engines that go beyond applying known techniques to standard problems
- Research into model interpretability, fairness, robustness, or adversarial resilience
- Creating new data preprocessing pipelines, feature engineering approaches, or augmentation strategies to overcome data limitations
- Developing reinforcement learning systems, generative models, or multi-agent architectures with uncertain convergence or performance
- Building ML infrastructure — custom training pipelines, inference optimization, model serving at scale — when standard approaches are insufficient
- Experimenting with transfer learning, few-shot learning, or domain adaptation techniques for novel applications

## SR&ED Rates for AI/ML Companies

- **CCPCs (most AI startups)**: 35% refundable credit on first $3M — AI companies often have high labour costs that drive large claims
- **Other corporations**: 15% non-refundable credit
- **Provincial top-ups**: Ontario (10%), Quebec (up to 30%), BC (10%), Alberta (8%)

GrantOps analyzes your GitHub commits, experiment tracking logs, and project tickets to automatically identify ML experimentation that qualifies for SR&ED.

- Automate your SR&ED claim: https://grantops.ai/en/automate-sred/
- Full SR&ED guide: https://grantops.ai/en/sred/
- Get started: https://app.grantops.ai