Affiliations: University of Waterloo & Vector Institute
Supervisors: Victor Zhong and Jimmy Lin
The David R. Cheriton School of Computer Science at the University of Waterloo and the Vector Institute invite applications for a Postdoctoral Fellow to lead a high-impact initiative on Agentic AI for Scientific Discovery.
This project aims to solve a grand challenge in modern AI: developing a deep research assistant capable of accelerating R&D workflows and operations in the natural sciences. While our initial exploration focuses on Chemistry, the core research addresses fundamental problems in how AI systems retrieve complex information, reason over it, and autonomously interact with proprietary scientific software. Crucially, this project places equal emphasis on methodology (building capable agents) and evaluation (defining how we measure agentic success). You will join a world-class ecosystem, combining the academic rigor of top-tier university labs with the resources and real-world data challenges of engaged industry partners.
About the Reading to Learn (R2L) Lab: The R2L Lab explores how language understanding improves machine learning efficiency and generalization. We are a leader in agentic benchmarks and evaluation; our platforms serve as primary evaluation standards for major industry labs like OpenAI, Anthropic, Google, and Salesforce. We focus on grounding, feedback, and synthetic data to build agents that reason and act effectively across digital and physical environments. For more information, please see our website.
The Postdoctoral Fellow will drive the research agenda, focusing on one or more of the following critical areas:
Multimodal Information Retrieval: Developing novel retrieval frameworks that unify heterogeneous scientific data (text, tables, molecular graphs, images, time series) drawn from massive data lakes.
Compositional Reasoning: Investigating new training methodologies that enable models to perform complex, multi-hop reasoning with strict provenance requirements.
Robust Agentic AI: Building generalist, multimodal agents (VLMs) capable of learning to operate complex, proprietary scientific software tools. This thrust emphasizes agent safety, error recovery, and tool-use.
Benchmarking, Simulation & Evaluation: The field lacks rigorous standards for evaluating agents in open-ended scientific tasks. You will lead the development of next-generation benchmarks—moving beyond static Q&A to dynamic, execution-based environments. This includes designing evaluation datasets, sandboxed scientific simulators, defining metrics for intermediate reasoning steps, and creating the definitive benchmarks for scientific agents.
Research Leadership: Lead a core research thrust and publish novel findings at top-tier AI conferences (e.g., NeurIPS, ICML, ICLR, ACL, SIGIR).
Defining Standards: Lead the design and release of open-source benchmarks and simulation environments that will define the standard for the AI for Science community.
Mentorship: Co-mentor a dedicated team of PhD and MSc students, fostering a collaborative lab culture.
Collaboration: Act as a bridge between academic PIs and industry partners, translating fundamental advances into deployed prototypes.
Education: A Ph.D. in Computer Science, Machine Learning, Chemistry, or a related field (completed or near completion). Note: While the application domain is Chemistry, we seek both strong AI/ML and Chemistry (organic chemistry preferred) researchers.
Track Record: A strong publication record in NLP, ML, Multimodality, Information Retrieval, or Chemistry.
Technical Skills: Proficiency in Python and PyTorch/JAX is required.
Research Focus: Experience with machine learning, reinforcement learning, VLMs, or the creation of rigorous datasets and evaluation protocols is highly preferred.
Term: Initial appointment is for one year, with the possibility of renewal for a years 2-3 based on performance.
Start Date: Flexible (Target: Summer/Fall 2026).
Compensation: Competitive salary (80,000-90,000 CAD).
Resources: Access to a project-dedicated H200 GPU cluster, travel funding for top-tier conferences, and the broader compute resources of the Vector Institute/Digital Research Alliance of Canada.
Please submit the following materials in a single PDF to Victor Zhong (vzhng@uwaterloo.ca) with the subject line "Postdoc Application: Agentic AI for Science":
Cover Letter: A brief (1 page max) statement outlining your research interests and specifying which of the research thrusts aligns best with your expertise.
Curriculum Vitae (CV): Including a list of publications and links to Google Scholar/GitHub.
In addition, please submit 2-3 recommendation letters emailed directly to vzhng@uwaterloo.ca. Review of applications will begin immediately and continue until the position is filled.
The University of Waterloo and the Vector Institute are committed to fostering an equitable and inclusive environment. We strongly encourage applications from all qualified individuals, including women, Indigenous peoples, persons with disabilities, and members of visible minorities.