Senior Research Engineer - ML Systems

Senior Research Engineer – ML Systems
Employment Type: Full-time
Company: Permute (www.permute.ai)

Permute is seeking a Senior Research Engineer to productionize, optimize, and extend the model systems that power AI reasoning over structured data. This role is for builders who can move from research ideas to reliable production systems, including the profiling, testing, and failure handling that prototypes often skip.

We care as much about how you think and build as we do about your background. The ideal candidate can implement research, diagnose model and systems performance, write clean production code, and make sound architectural decisions in a fast-moving startup environment.

• Productionize and optimize our existing learned evidence architecture for structured data

• Improve training and inference performance, including throughput, latency, memory use, reliability, and cost

• Port and optimize model training and inference workloads from CPU to GPU

• Build production systems supporting model training, evaluation, deployment, and inference

• Develop tooling for experimentation, reproducibility, monitoring, and observability

• Write clean, maintainable Python and PyTorch systems that integrate with Permute's broader platform

• Design and evaluate new heads, layers, objectives, and fine-tuning methods

• Explore new model variants, including transformer-based architectures and reinforcement learning

• Collaborate with engineering and product teams to deliver model capabilities that power production AI features

• Strong background in machine learning research and ML systems

• Experience building and training models with PyTorch

• Strong foundation in algorithms, statistics, optimization, and experimental design

• Strong software engineering and system architecture skills

• 5+ years building ML or performance-sensitive software systems

• Degree in Mathematics, Physics, Computer Science, or a related technical field

Experience with:

• End-to-end production ML systems

• Model training, MLOps, evaluation, and deployment

• Performance engineering, including CUDA, Triton, quantization, or model compilation

• Transformers, fine-tuning, post-training, or reinforcement learning

• Meaningful contributions to open-source ML frameworks or model implementations