Role Summary

Company: Lexsi Labs

Location: Remote

Duration: 6 Months (Potential Extension / Full-Time Conversion)

Stipend: Competitive

Start Date: Rolling

About Lexsi Labs

Lexsi Labs is a frontier AI research lab focused on building aligned, interpretable, and safe superintelligence. The team works on advancing AI alignment, mechanistic interpretability, uncertainty estimation, reinforcement learning, and tabular foundation models to create reliable and transparent AI systems.

As an AI Research Intern, you will work alongside researchers and engineers on cutting-edge AI challenges, contribute to novel research initiatives, and help build next-generation AI technologies that prioritize safety, transparency, and performance.

Roles & Responsibilities

Depending on your interests and expertise, you will contribute to one or more of the following areas:

AI Research & Development

• Design, implement, and evaluate novel AI research ideas.
• Build and maintain experimental pipelines and research infrastructure.
• Conduct systematic experiments and analyze research outcomes.

Explainability & Interpretability

• Apply and improve Explainable AI (XAI) techniques such as SHAP, LRP, Grad-CAM, Integrated Gradients, and related methods.
• Investigate model behavior across text, image, and tabular domains.
• Develop insights into model decision-making and trustworthiness.

Mechanistic Interpretability

• Explore internal representations and circuits within large AI models.
• Use activation patching, feature visualization, and related techniques to understand emergent behaviors and failure modes.

Uncertainty & Robustness

• Develop uncertainty estimation and robustness evaluation methods.
• Benchmark Bayesian approaches, ensemble techniques, and test-time augmentation strategies.
• Assess AI model reliability in real-world scenarios.

Foundation Models & LLM Research

• Work with Transformer architectures including GPT, BERT, LLaMA, T5, MoE, and Mamba-based models.
• Explore model alignment, post-training adaptation, fine-tuning, and instruction tuning techniques.
• Contribute to reinforcement learning and RLHF-related research initiatives.

Open-Source & Library Development

• Develop and improve Python libraries for AI alignment, interpretability, robustness, and machine unlearning.
• Build reusable tooling and research frameworks for the AI community.

Research Publications

• Document experiments and findings.
• Contribute to technical reports, whitepapers, research papers, and conference submissions.

Eligibility Criteria

Required Skills

• Strong proficiency in Python and software development best practices.
• Solid understanding of Machine Learning and Deep Learning fundamentals.
• Hands-on experience with PyTorch.
• Strong understanding of Transformer architectures, attention mechanisms, tokenization, positional encodings, and training objectives.
• Familiarity with Git, version control workflows, and collaborative software development.
• Strong analytical thinking and research mindset.

Preferred Skills (Any One Area)

Explainable AI (XAI)

• Experience with SHAP, LIME, Integrated Gradients, LRP, Grad-CAM, or related techniques.

Mechanistic Interpretability

• Understanding of circuit analysis, activation patching, and feature visualization.

Uncertainty Estimation

• Experience with Bayesian methods, ensemble models, or uncertainty quantification techniques.

Model Optimization

• Knowledge of quantization, pruning, and model compression.

LLM Alignment & Fine-Tuning

• Experience with RLHF, reward modeling, instruction tuning, LoRA, adapters, knowledge distillation, or domain adaptation.

Tabular Foundation Models

• Familiarity with frameworks such as Orion, TabPFN, TabICL, or related architectures.

Preferred Qualifications

• Research publications in top-tier AI conferences or journals.
• Contributions to open-source AI/ML projects.
• Experience working on large-scale AI systems.
• Exposure to domains such as healthcare, finance, or other high-impact applications.
• Familiarity with distributed training and performance optimization techniques.

Perks & Benefits

• Work on cutting-edge AI research problems.
• Access to GPUs, cloud infrastructure, and advanced AI models.
• Competitive stipend.
• Opportunity for full-time conversion based on performance.
• Mentorship from experienced AI researchers and engineers.
• Co-authorship opportunities on research papers, technical reports, and conference submissions.
• Exposure to frontier research in AI alignment, interpretability, and foundation models.

Why Join?

This internship offers a unique opportunity to work at the intersection of research and innovation, contribute to breakthrough AI technologies, and collaborate with a team dedicated to building safe, transparent, and impactful AI systems.