The Lexsi Lab is a premier research lab dedicated to solving one of the most critical challenges of our time: ensuring that advanced artificial intelligence systems are safe, transparent, and aligned with human values. Founded by pioneers in the deep learning space, our lab operates at the global forefront of AI research with hubs in Mumbai and Paris. We tackle foundational problems in explainability, robustness, and reasoning, publishing our work in top-tier venues (ICML, IJCNN, WWW, ACL) and creating patented, open-source tools like DL-Backtrace that push the boundaries of the field.

Key areas where we are looking to expand our R&D: New RL Techniques for efficient and faster learning; New Tabular Foundational Models; Interetability Alignment Strategies like Pruning, Unlearning, Safety fine-tuning, and new exploration agents.

We are a team of world-class researchers, engineers, and visionaries, seeking passionate individuals to join us in shaping a future where AI operates as a beneficial and trustworthy partner to humanity.

The Role

As an AI Research Scientist, you will work at the heart of our research agenda, tackling fundamental questions in AI alignment and interpretability. You will have the freedom to pursue ambitious, long-term research projects while collaborating with a team of leading experts.

Your responsibilities will include:

• Conducting foundational research in areas such as mechanistic interpretability, explainability, fairness-aware model design, robustness, and foundational reasoning models.
• Developing and testing novel algorithms and techniques to understand, interpret, and control the behavior of large-scale AI models (LLMs, MLLMs, Diffusion Models).
• Investigating and mitigating failure modes, such as data memorization, bias, and model hallucinations.
• Contributing to the development of our in-house foundation models for high-stakes domains like finance and healthcare.
• Collaborating with our global team to publish cutting-edge research in top-tier AI conferences and journals.
• Engaging with the broader AI safety community and contributing to the open-source tools we develop.

Who You Are

You are an inquisitive and driven researcher with a passion for solving the AI alignment problem. You are motivated by challenging questions and thrive in a collaborative, intellectually stimulating environment.

Ideal Qualifications:

• A PhD, or a Master's degree with equivalent research experience, in Computer Science, AI, Machine Learning, or a related quantitative field.
• A strong publication record in leading AI conferences (e.g., NeurIPS, ICML, ICLR, CVPR, ACL, AAAI, etc.).
• Deep technical knowledge of modern machine learning, especially deep learning architectures like Transformers, GNNs, and generative models.
• Hands-on experience building and training large models using frameworks like PyTorch or TensorFlow.
• Excellent programming skills, particularly in Python.
• A demonstrated interest and passion for AI safety, ethics, and interpretability.

What We Offer

• A Mission-Driven Culture: An opportunity to work on what we believe is the most important and impactful challenge of the 21st century.
• Intellectual Freedom: The resources and support to pursue ambitious, high-impact research directions.

Join us in building a safe & aligned AI for the future.

To Apply: Please send your CV, a link to your Google Scholar profile, and a brief statement of your research interests.