Saeid Asgari

I am a Principal Applied Science Manager at Microsoft US (previously Principal Researcher) and an Adjunct CS Research Professor at Simon Fraser University. Our team at Microsoft pushes the boundries of LLM evaluations at scale. Previously, I worked on reliable AI at Autodesk Research and explored ML/CV problems in startups.

I completed my PhD in Computing Science at SFU under the supervision of Prof. Ghassan Hamarneh. During my PhD I had multiple research visits/internships at different research institutions such as MILA (Montreal, Canada) and Siemens Healtheneers (Princeton, US).

Outside of work, I have a strong passion for soccer and running track - the 100m sprint. I've had the privilege of competing and earning a few medals in inter-university competitions in these areas.

I lead research and development initiatives in reliability and evaluation of generative models like LLMs and VLMs. I am broadly intrested in:

• Reliability and Robustness: Designing experiments and datasets to analyze generalization, memorization, and robustness in generative models, ensuring they perform reliably across diverse scenarios.
• Interpretability and Explainability: Building tools to understand ML model behavior, improve transparency, and ensure models meet safety and ethical standards.
• Data Bias and Mitigation: Developing scalable tools to identify hidden and known biases in datasets, while enhancing datasets and designing models to better handle biases, out-of-domain samples, and improve generalization.

Design template by Jon Barron