Shaina Raza, PhD

HumaniBench: Human-Centric Evaluation for Multimodal Models

Link: HumaniBench

Summary: A new benchmark to evaluate fairness, social bias, and empathy in large multimodal models (VLMs, LMMs). Includes multilingual and adversarial test cases.

FairSense-AI: Fairness and Sustainability Toolkit

Institute: Vector Institute

Link: FairSense-AI Project

Summary: Developed a comprehensive framework to evaluate and improve fairness, sustainability, and explainability in large-scale AI systems. Includes methods for auditing bias, reducing emissions, and ensuring compliance with responsible AI principles.

VLBench: Benchmarking Vision-Language Models

Link: VLBench

Summary: Introduced a novel benchmark for evaluating robustness and alignment in vision-language models, focusing on fairness and multilingual support.

UnBias: Framework for Bias Detection and Mitigation

Summary: Developed a pipeline to surface, audit, and reduce bias in generative and classification models using social cues. Includes benchmark data and bias metrics across language, vision, and structured inputs.

Documentation: UnBias ReadTheDocs

Equity in Public Health AI Systems

Institute: Ontario Agency for Health Protection and Promotion (OAHPP)

Summary: Focused on embedding fairness, demographic transparency, and equitable decision-making in AI systems deployed in public health surveillance and intervention planning.

Health System Impact (HSI) Fellowship

Funding: CIHR & Ontario Agency for Health Protection and Promotion (OAHPP)

Link: CIHR HSI Fellowship

Summary: One of the select national recipients of the HSI fellowship awarded for impactful research in responsible AI and equitable public health data systems.

Green AI Optimization

Summary: Multiple publications focused on sustainability, quantization, and model efficiency:

• Reducing Emissions in LLMs
• Sustainable AI Practices
• Benchmarking Energy Footprints

Model Immunization

Link: Just as Humans Need Vaccines

Summary: Introduced techniques for immunizing LLMs against falsehoods and harmful content, improving robustness and factuality.

Keynotes and Talks

• AI Symposium at Conestoga College (Nov 2, 2024): "AI and Data Privacy and Ethics" – Keynote on integrating ethical practices in AI development.
• Future Votes – Cybersecurity & Election Integrity (Oct 24, 2024): "The Role of AI in Securing Elections" – Impact of AI on election integrity.
• Fairness in Public Health and Equity (York University, Nov 5, 2024): "Enhancing Equity in Healthcare Through AI" – Promoting equitable outcomes.
• Mitigating Biases in Text – LLMs (Cohere, USA, July 2024): "Challenges and Strategies" – Developing less biased LLMs.
• Cybersecurity and AI – Who Loses? (Rogers Cybersecure Catalyst, Jan 2024): "Balancing Cybersecurity and Inclusivity" – Advocating inclusive AI.
• Addressing News Media Biases (AI.Science Webinar, Mar 8, 2024): "Combating Misinformation" – Tools to counteract news bias.

Peer Review Committee Member – CIHR Doctoral Research Awards

Role: Peer Review Committee Member

Link: CIHR Doctoral Review Committee

Summary: Participated as a reviewer for Canada’s national CIHR Doctoral Awards program, evaluating top-tier PhD proposals in health-related AI and data science fields.

Peer Reviewer – National Killam Program

Role: Peer Reviewer

Summary: Reviewed nominations for Canada’s prestigious Killam Prizes and Fellowships, recognizing excellence in research across disciplines.