Role overview
Classie is an early stage startup at the intersection of product and research. Team members have deployed at scale in companies such as Google, Nvidia, VMware, have academic background from Stanford, and have done over 20 startups among them. Classie is building the first platform to supervise AI in real-time, using a systems-first approach to explain and secure complex AI agent behavior at high token volume.
We're looking for a ML Systems Engineer who will help design, build and maintain our rapidly expanding core ML systems. You’ll solve hard problems working alongside a team of cross-stack experts who have done it before, contributing to the creation of a category-defining company.
**Responsibilities**
* Design and implement ML inference machinery for fast, cost-effective, and high-accuracy real-time analysis.
* Develop and maintain data pipelines for activity tracing, storage, and event detection across the platform.
* Build and scale training infrastructure, and develop self-learning mechanisms.
**Qualifications**
Required: Deep knowledge of PyTorch or JAX
What we're looking for
* Designing data synthesis pipelines and dataset construction.
* Designing evaluation frameworks and custom benchmarks to measure model capability
* Implementing model acceleration techniques (e.g., Sparsification / Quantization-aware training)
* Working knowledge of high-throughput serving systems (e.g., vLLM / SGLang)
* Working knowledge of RL systems (e.g., verl) is a plus.