Research Engineer - Model Evaluation & MLOps

Sciforium•San Francisco, CA

10d•$155,000 - $200,000

About The Position

As a Research Engineer focused on Model Evaluation & MLOps, you will build the tools and infrastructure needed to evaluate, deploy, and operate multimodal foundation models reliably. You will rapidly enable Sciforium’s models and the latest open-weight models on GPUs, automate quality and performance benchmarking, and improve the MLOps workflows that connect research experiments to reliable releases.

Requirements

2+ years of professional ML or software engineering experience, including work on production ML systems, ML platforms, or MLOps infrastructure.
Strong Python and software engineering skills, with experience building reliable production systems.
Hands-on experience with PyTorch, TensorFlow, or JAX and a good understanding of modern language or multimodal model architectures.
Experience with model evaluation or benchmarking and core model lifecycle workflows such as experiment tracking, versioning, deployment, or monitoring.
Experience running, benchmarking, and debugging models with one or more GPU inference runtimes, such as vLLM, SGLang, TensorRT-LLM, or equivalent, in containerized cloud or on-premises environments.
Ability to document systems clearly and collaborate across research, infrastructure, and product engineering teams.
MS or PhD in Computer Science, Computer Engineering, Machine Learning, or a related technical field, or equivalent practical experience.

Nice To Haves

Familiarity with Hugging Face Transformers or similar model libraries.
Experience enabling models on AMD GPUs and ROCm.
Contributions to open-source evaluation, model, or ML infrastructure projects.

Responsibilities

Rapidly integrate new internal and open-weight language and multimodal models into our GPU evaluation and inference environments.
Build automated benchmarks for model quality and systems performance, including latency, throughput, and memory usage.
Create standardized, reproducible comparisons across Sciforium models, external baselines, and runtime configurations.
Build and maintain experiment tracking, model registry, and versioning for models, datasets, and evaluation configurations.
Automate the path from research checkpoints to validated deployments through CI/CD and reproducible workflows.
Monitor model quality and systems performance, and diagnose failures or regressions across model and deployment pipelines.
Build reusable tools that help researchers launch evaluations, compare experiments, and reproduce results.
Profile end-to-end model workloads and collaborate with distributed systems, inference, and GPU kernel engineers on deeper performance issues.