Job description
KITCHI is N51's cognitive decision-support layer, the AI system that reads the fused picture INUKSHUK builds and surfaces ranked courses of action with confidence scores and full reasoning visible to the commander. Explainability is not an afterthought in KITCHI, it is the core product requirement. DND/CAF evaluators and Challenge 13 reviewers will assess the quality and trustworthiness of KITCHI's outputs directly. As ML Engineer for Explainable AI, you will own the models, the inference pipeline, and the XAI output layer that makes KITCHI's reasoning legible to operators in real time.
What will you do?
Design, train, and evaluate ML models for multi-domain threat assessment, anomaly detection, and course-of-action generation from fused sensor data. Implement explainability frameworks, SHAP, LIME, attention visualisation, or custom approaches, that surface decision rationale in operator-readable formats with appropriate confidence scoring and uncertainty quantification. Build the inference pipeline for edge deployment: model compression, quantisation, and optimisation for <100ms end-to-end latency on tactical hardware. Work with the sensor fusion team to ensure KITCHI's inputs are correctly structured and with the platform team to ensure its outputs are correctly displayed. Contribute to XAI methodology documentation for DND/CAF technical evaluation.
Requirements:
- 5+ years in ML engineering with production model deployment experience
- Deep familiarity with explainability methods — SHAP, LIME, integrated gradients, or equivalent
- Experience with uncertainty quantification and calibration in deployed ML systems
- Strong skills in model optimisation for edge inference — quantisation, pruning, distillation
- Experience with time-series or spatiotemporal data is a strong asset
- Familiarity with defence decision-making frameworks or human-machine teaming research is a strong asset
- Canadian citizenship or permanent residency required
- Security clearance eligibility required