People will soon realize that automated prompt engineering is akin to latent variable inference. Then they will realize that latent variable inference is akin to reasoning/planning. Then they will realize that transformers were missing that piece all along.

Exciting news! Our recent collaborative work with TELUS, Deloitte Canada & Roche Canada, “Using Social Media to Help Understand Patient-Reported Health Outcomes of Post–COVID-19 Condition: NLP Approach” has been recently published in JMIR Publications jmir.org/2023/1/e45767

Exciting Applied ML opportunity for Biomarker Discovery: drive.google.com/file/d/1w_UMjK…

Ambient Intelligence 4 Pregnancy Monitoring: journals.plos.org/digitalhealth/…

On #WorldMentalHealthDay, we couldn't agree more with Elham Dolatabadi on AI's potential to support youth mental health.🧠💙 Collaborating with Kids Help Phone we’ve developed the FAIIR tool to help frontline staff deliver human-centred care to youth. The FAIIR AI tool uses

#YorkU Prof Elham Dolatabadi and team develop an AI tool to monitor sleep position for pregnant women to see if back sleeping or other postures effect the health of their unborn child YorkUHealth Elham Dolatabadi yorku.ca/news/2024/12/0…

NSERC CRC Tier 2 position at York University. Excellent opportunity to work with internationally renowned researchers and top talent: yorku.ca/vpepc/faculty-…

seeking a highly motivated Postdoctoral Research Fellow to join my lab (multimodal learning, multi-omics representation learning, fairness, responsible AI). Please send an email to [email protected] with the subject "Postdoc Application."

My first PhD paper: VLM & ViT inference speedup 👇 (a thread) Why stick to simple attention ranking for token pruning? Our paper "Similarity Aware Token Pruning: Your VLM but Faster" tackles both ViTs & VLMs with a dynamic, training-free approach for SOTA efficiency/performance.

happy to introduce 'SAINT': arxiv.org/abs/2503.11549 SAINT sets a new benchmark, achieving SOTA performance-efficiency trade-off on ViTs ...; We just dropped "Similarity Aware Token Pruning:

Advancing medical AI through high-quality data: Introducing Open-PMC Vector AI Engineering Team Members Arash Afkanpour, Adibvafa Fallahpour, Franklin O., Shuvendu Roy, Sajad Ashkezari, and Vahid Reza Khazaie, along with Vector Faculty Affiliate Elham Dolatabadi, & Vector

Excited to announce the release of Open-PMC, a high-quality medical image-caption dataset! This dataset is designed to advance research in multimodal AI for healthcare. Huge thanks to the incredible team who worked tirelessly on this for months. arxiv.org/abs/2503.14377

New research finds specific learning strategies can enhance AI model effectiveness in hospitals yorku.ca/news/2025/06/0…

How can we teach massive vision models new tasks with just a few examples, without forgetting what they already know? Check our recent paper, published at TMLR: lnkd.in/gHc49iBw

History doesn’t just repeat itself, it accelerates.

The American people voted for peace. Remember that?

#no war instagram.com/reel/DK2H2x5Ry…

springwise.com/healthy-societ…