Fully Funded Postdoctoral Researcher in Trustworthy AI Assistants for Scientific Research – Mila / ÉTS, Montréal, Canada

Position Snapshot

• Institutions: Mila – Quebec AI Institute, and École de technologie supérieure (ÉTS), Montréal, Canada

• Supervision: Hugo Larochelle (Mila) and Ulrich Aïvodji (ÉTS/Mila)

• Level: Postdoctoral Researcher (full-time)

• Location: Montréal, Canada

• Duration: 24 months

• Research areas:

  • AI-assisted peer review and evaluation of AI reviewers

  • Verifiable mathematical reasoning for AI/ML

  • LLM evaluation, fairness, and benchmarking

  • Human-AI collaboration for scientific workflows

• Application deadline: 15 September (applications reviewed as soon as possible, no later than this date)

• How to apply: Email application materials to ulrich.aivodji@etsmtl.ca and hugo.larochelle@mila.quebec

About the Position

Hugo Larochelle and Ulrich Aïvodji are recruiting a postdoctoral researcher for a 24-month project on trustworthy AI assistants for scientific research, based in Montréal.

Artificial agents equipped with natural language processing and tool-use capabilities are rapidly changing the daily work of AI researchers — summarizing papers, generating hypotheses, assisting with coding, drafting reviews, polishing manuscripts, checking mathematical arguments, and supporting scientific decision-making. These capabilities could help researchers work faster and verify claims more rigorously, but they also raise risks: hidden biases, homogenized feedback, paper laundering, and unverifiable claims that could weaken rather than strengthen the research ecosystem.

The goal of this postdoctoral project is to study how AI systems can support the work of AI researchers while preserving scientific rigor, fairness, diversity of perspectives, and verifiability. The project centers on two complementary work streams.

1. Evaluating AI-Assisted Peer-Review Systems

This stream studies AI reviewers and AI-assisted reviewing systems as high-stakes decision-support tools. It asks which parts of the peer-review pipeline can be safely assisted by AI, under what constraints, and with which evaluation criteria — building on recent findings on affiliation bias, hidden status-based preferences, excessive agreement, and vulnerability to stylistic manipulation in AI reviewers. The work will also explore more constructive designs, including reviewer personas, ensembles, metareviewing, literature-grounded reviewing, and human-AI collaboration protocols.

2. Developing AI Assistants for Verifiable Mathematical Reasoning

This stream studies the use of LLMs to formalize typical mathematical claims in AI/ML papers, identifying which classes of claims are currently amenable to AI-assisted theorem proving and which remain beyond the reach of existing systems. It will also examine how retrieval from formal libraries can improve proof search in domain-specific AI/ML settings, and how verification-aware research assistants can help authors catch missing assumptions, ambiguous theorem statements, and reasoning gaps before submission.

Setting and Environment

The postdoc will join a research environment focused on trustworthy AI, privacy, fairness, and explainability, benefiting from connections with Mila, ÉTS, and the broader research community working on AI evaluation and governance. The position offers significant freedom to shape the research direction — a successful candidate may lean more strongly toward AI reviewer evaluation, formal verification for AI/ML mathematics, or the intersection of both. Across both directions, the project emphasizes open science, reproducible evaluation, and tools useful to the research community.

Required Qualifications

Highly motivated candidate with a strong interest in trustworthy AI and AI-assisted scientific research, particularly with:

• A strong background in machine learning, natural language processing, trustworthy AI, formal methods, or AI for science

• Good programming skills in Python and modern machine-learning frameworks

• Strong written communication skills and the ability to formulate research ideas clearly

• A PhD in computer science, machine learning, or a related field — completed or close to completion

Valuable Skills (Assets)

• Experience with LLM evaluation, benchmarking, fairness, or peer review systems

• Experience with theorem proving, Lean, formal verification, or mathematical reasoning

Application Process

Applications should be sent by email to ulrich.aivodji@etsmtl.ca and hugo.larochelle@mila.quebec, as soon as possible and no later than September 15th, and should include:

1. Detailed CV

2. Letter of motivation

3. Two references or recommendation letters

Why This Opportunity Stands Out

This postdoctoral position offers a rare combination of:

• Timely, high-impact research on trustworthy AI at the intersection of peer review, LLM evaluation, and formal mathematical reasoning

• Joint supervision by Hugo Larochelle (Scientific Director of Mila, one of the world's leading deep learning research centers) and Ulrich Aïvodji (Trustworthy Information Systems Lab, ÉTS)

• Significant research freedom, with the option to focus on AI reviewer evaluation, verifiable mathematical reasoning, or both

• Deep connections to Mila's ecosystem of 1,500+ researchers and 150+ industrial partners, and the broader Montréal AI community

• A strong emphasis on open science and reproducibility, producing tools of direct use to the research community

For candidates passionate about trustworthy AI, LLM evaluation, and the future of scientific workflows, this is an exceptional opportunity to pursue high-impact postdoctoral research in one of the world's leading AI hubs.

Details & Application Instructions

Full proposal (PDF): https://aivodji.github.io/positions/pdfs/AI-Assistant-Postdoc_proposal.pdf

For more personalized tips or templates contact our team.

www.applywithmentors.com