Machine Learning Engineer: LLM Interpretability & Systems

What you'll do

• Translate mechanistic interpretability research into production code and improve commercial and open-source models using their internals.
• Use activation patching, control vectors, and feature extraction to make targeted, repeatable improvements to model output.
• Build evaluation, deployment, and feature-level intervention systems for reliable policy enforcement at inference time.

What they ask for

• Strong understanding of Transformer architectures, PyTorch internals, and the mathematical foundations of deep learning.
• Experience training, fine-tuning, or optimizing models beyond superficial augmentation.
• Ability to interpret research papers and implement relevant methods.

Pay

$175K – $250K (base salary)

Tools

Python, Rust, Node/TypeScript, React, Postgresql, Docker, Kubernetes, Terraform

AI focus

Mechanistic interpretability and model-internal interventions to improve LLM performance, reliability, controllability, and deterministic policy enforcement.