Machine Learning Engineer at CTGT, Inc.

You will operate deep within the model stack to build the deterministic governance layer for enterprise AI. By leveraging mechanistic interpretability, you'll work directly with model internals—weights and activations—to enforce policy and prevent drift. This role transforms frontier research into production systems that make LLMs reliable for Fortune 500 institutions.

• Implement techniques like activation patching and control vectors to achieve targeted, repeatable improvements in model output.
• Design and optimize feature-level intervention systems that enable deterministic policy enforcement at inference time for commercial and open-source models.
• Build the evaluation and deployment loops required to ship interpretability-based changes reliably into complex enterprise environments.

• Possesses a deep mathematical foundation in Transformer architectures and PyTorch internals, with experience training or fine-tuning models beyond superficial augmentation.
• Demonstrates the ability to translate academic papers on mechanistic interpretability into robust, production-ready code.
• Exhibits an ownership mindset and technical curiosity, driven to solve the challenge of making non-deterministic models auditable and controllable.

• Work at the intersection of frontier AI research and production environments, moving beyond simple prompting to influence the mechanics of model cognition.
• Join a high-pedigree team born out of Stanford research, backed by elite investors including Google’s Gradient Ventures and Y Combinator.
• Drive massive impact by building the core "Policy Engine" that enables the world's most important institutions to deploy generative AI with confidence.