Job Details

Title: AI Researcher

Thoth AI · Bay Area, CA (Hybrid)

About Thoth AI

Thoth AI builds data infrastructure for the AI industry. Our customers — robotics companies, foundation model teams, and enterprise AIorganizations — rely on us to provide the high-quality, multimodal training data their models need to work in the real world. We're now building a Bay Area R&D team to develop the next generation of our core technology.

The Role

We're looking for AI Researchers to join a small, high-impact R&D team working on applied research problems across multimodal learning, LLM reliability, and intelligent data systems. You'll work on projects that include hallucination detection algorithms, LLM gating and output verification, active learning systems, sim-to-real domain transfer, and generative approaches to data — with a direct line from research to production.

This is a founding R&D hire. You'll shape the research agenda, not just execute it.

What You'll Work On

• Design and build hallucination detection systems — algorithms that identify when LLMs produce unreliable outputs, using approaches like internal activation probing, token-level entropy analysis, and semantic consistency checking
• Develop LLM gating and output verification methods that decide in real time whether a model's response should be trusted, flagged, or blocked
• Build active learning pipelines that analyze model failure patterns and automatically recommend what data to collect next, how much, and in what distribution
• Research sim-to-real domain transfer and generative techniques that bridge the gap between synthetic and real-world data across visual, spatial, and sensor modalities
• Collaborate with engineering and product teams to translate research into scalable platform capabilities
• Publish and present work at top-tier venues when it advances the company's mission and your career

What We're Looking For

• MS or PhD in Computer Science, Machine Learning, Robotics, or a related field
• Strong foundations in deep learning, with depth in one or more of: LLM internals and evaluation, hallucination detection, generative models, representation learning, domain adaptation, sim-to-real transfer, active learning, or multimodal learning
• Proficiency in Python and modern ML frameworks (PyTorch preferred)
• Experience working with complex, real-world data — video, 3D, language, sensor, or a combination
• Track record of research output (publications, open-source contributions, or shipped ML systems)
• Ability to operate independently in an early-stage environment while collaborating closely with a small team

Nice to Have

• Experience with token probability analysis, attention probing, or other white-box/gray-box LLM inspection methods
• Familiarity with RLHF, preference learning, or LLM alignment techniques
• Experience with robotics data, egocentric video, or embodied AI
• Familiarity with simulation environments (NVIDIA Isaac Sim, Unity, Unreal Engine)
• Prior experience at a startup or in applied research bridging academia and production