Research
The research trail, grouped by theme and newest first inside each group. It starts with the question most of it shares: how do you check what a model tells you?
Interpretability and evaluation
Apr 2024
Single-author literature review
Literature review
Interpretability and Explainability in Generative AI
Reviewed what interpretability and explainability research offers against generative AI's failure modes (hallucination, misgeneralization, bias, prompt injection, deepfakes), drawing on 14papers in the literature review's reference listThe reference list on the review's last two pages; some are surveys cited for context.Source: Literature review, 21 April 2024 papers, from model-graded resilience scoring and causal-inference benchmarking to saliency-aware counterfactuals and rule extraction from reinforcement-learning agents. Concluded that no method was yet widely deployed, and that standardized interpretability metrics weighted by application domain are essential to deploying these systems with confidence.
- Generative AI
- Interpretability (XAI)
Apr – Aug 2022
SFU Engineering Science | Supervisor: Dr. Ivan V. Bajić (SFU Multimedia Lab)
Honours thesis
Evaluation of Feature Attribution Methods in Interpretable Machine Learning
Applied Zhou et al.'s induced-ground-truth design (AAAI 2022) to chest X-rays: watermarked them, then used IoU and paired t-tests to measure how precisely fiveclass activation map (CAM) explanation methods comparedSource: Resume, revised 2026-09-11 CAM methods localized the watermark; found FullGrad the most robust and precise, improved further by thresholding.
- PyTorch
- Interpretability (XAI)
- Medical imaging
- CNNs
Sept 2021 – May 2022
Stryker R&D, Computer Vision
Industry research
Explainable AI for surgical video analysis
Applied and evaluated explainable AI (XAI) techniques in PyTorch — including saliency mapping, feature attribution, and counterfactual reasoning — to improve model transparency and support surgeon decision-making.
- PyTorch
- Interpretability (XAI)
- Computer vision
Medical and neural AI
Jan – Aug 2022
SFU Engineering Capstone; Best Overall Project, ICAMES 2022
Capstone
Adolescent Spinal Curvature Brace Optimization
A wearable pressure-sensing system (force-sensor array, noise-filtering firmware, real-time pressure heatmap on a CAD torso model) to guide orthopedic brace design, following medical-device software standards (IEC 62304).
- Firmware & sensors
Aug 2020 – Apr 2021
Rostrum Medical Innovations Inc.; HealthTech Connex Inc. (BrainNet)
Research role
Researcher – Rostrum Medical Innovations & HealthTech Connex
Researched capnography classification and EEG-based disease detection with interpretable multimodal deep learning for explainable medical AI.
- Biomedical signals (EEG, capnography)
- Deep learning
- Interpretability (XAI)
May – Aug 2018
SFU Functional & Anatomical Imaging & Shape Analysis (FAISAL) Lab
Research role
Medical Imaging Research Assistant
Processed and analyzed CT scan data, applying segmentation correction techniques relevant to computer vision.
- Medical imaging
- Segmentation
Applied LLM systems
Sep 2023
Microsoft Global AI Hackathon
Hackathon
Clinical CoPilot
Built a retrieval-augmented generation (RAG) pipeline for a clinical LLM using Azure OpenAI and Semantic Kernel; ranked 16th/1,253team ranking at Microsoft's Global AI HackathonExecutive Challenge honourable mention (LinkedIn).Source: Resume, revised 2026-09-11 teams with an Executive Challenge honourable mention at Microsoft's Global AI Hackathon.
- RAG
- Azure OpenAI
- Semantic Kernel
Reviewing
Conference work, also listed under leadership.
Jun 2024
Microsoft Machine Learning, AI & Data Science Conference
MLADS Co-Area Chair & Proposal Reviewer
As co-area chair & proposal reviewer for Microsoft's MLADS AI/ML conference, evaluated and selected talks and posters in forecasting, anomaly detection, and semi-supervised learning.