About the role#
Frontier models are often right but occasionally wrong, and they cannot distinguish between the two. This is a problem for the high-stakes environments we serve. Our research team works to bridge this gap by using feature learning to extract and steer model behavior at runtime. As a research intern, you will own a specific problem from start to finish. You will not be given busy work. You will define an approach, implement it against real models, and provide evidence that holds up under scrutiny. You will work alongside engineers to build capabilities for our Policy Engine and participate in weekly research reviews where you must defend your reasoning.
What you'll do#
- Implement and stress-test methods for feature extraction and runtime intervention.
- Design evaluations that bound error rather than average it.
- Read relevant literature, determine what matters, reproduce findings, and push past them.
- Work with engineering to turn research findings into Policy Engine capabilities that ship into audited environments.
- Present your progress weekly and defend your technical reasoning.
What you'll need#
- Enrollment in a degree program (Bachelor's through PhD) in computer science, mathematics, physics, or a similar quantitative field.
- Strong mathematical foundations in linear algebra, probability, optimization, and information theory.
- The ability to read research papers, identify key insights, and implement them.
- Experience writing real code for computational systems.
- Fluency with PyTorch and the modern ML stack, or a proven track record of picking up new tools quickly.
- An interest in interpretability, model internals, and making systems provably reliable.
- A self-directed approach to work, capable of making progress without constant supervision.
Location & details#
- This is a full-time, on-site internship based in San Francisco, California.
- The term is for Summer 2027.
- We provide visa sponsorship for this role.
About CTGT
CTGT is a research services firm based in San Francisco. Founded in 2023, the company provides representation-level control for artificial intelligence systems. It focuses on delivering audit trails and mathematical certainty for large enterprises. The team consists of 13 employees.
How to get in at CTGT
Securing a role at CTGT requires speed because early applicants often get their materials reviewed before the candidate pool grows. Intern Insider sends an instant alert the moment a role matching your target is published, which helps you apply among the first and gain a visibility advantage. Applying early is a practical way to ensure your profile is seen by the team. You can also improve your response rates by reaching out to the recruiters behind CTGT roles directly. Intern Insider surfaces these specific contacts so you can ask informed questions about the role or request a referral. This direct approach is often more effective than submitting an application into a general queue.
