Member of ML Technical Staff

Pragmatike•San Francisco, CA

3d•$200,000 - $350,000•Onsite

About The Position

We are looking for an exceptional Member of Technical Staff specializing in Machine Learning and Large Language Models to join an early-stage AI company building and training state-of-the-art foundation models. This role sits at the intersection of LLM research, large-scale training infrastructure, post-training, and GPU/kernel optimization. We are particularly interested in highly motivated researchers and engineers who want to contribute directly to training powerful models — whether their strengths are in theoretical model research, training systems, distributed infrastructure, or low-level performance optimization. You will work in a small, highly technical team where researchers and engineers collaborate closely and are expected to take ownership across the stack.

Requirements

At least 1+ years of experience in theoretical LLM research or as an ML researcher/engineer at a highly technical AI or technology organization.
Hands-on experience working with large language models beyond simply consuming existing APIs.
Experience with one or more of: LLM architecture research, Pre-training, Post-training, Reinforcement learning / preference optimization, Training framework development, Kernel or inference optimization, Large-scale distributed training.
Experience working on language models at organizations or research environments comparable to OpenAI, Google DeepMind, Mistral AI, Qwen, DeepSeek, Z.ai, Allen Institute for AI, or leading academic labs is highly relevant.
Strong understanding of large-scale AI infrastructure and at least some of the following: Distributed GPU training, Model parallelism, Tensor parallelism, Pipeline parallelism, Sequence parallelism, Data parallelism, Communication optimization, Memory optimization, Training throughput optimization, Software/hardware co-design.
Strong proficiency with: Python, PyTorch, CUDA, Triton.
We value candidates who have demonstrated strong technical depth through one or more of: ML/AI research during undergraduate, master's, or PhD studies, Publications or meaningful research contributions, Open-source ML contributions, Competitive programming, Building large-scale ML systems from first principles.
A strong undergraduate degree is expected, ideally from a highly selective technical university.

Nice To Haves

Experience at an early-stage AI startup.
Contributions to open-source ML frameworks or research projects.
Experience optimizing GPU kernels or inference engines.
Experience building training infrastructure from scratch.
Experience training models across large GPU clusters.
Strong systems engineering or HPC background.
Experience with JAX is highly valued.
Additional experience with distributed training frameworks, custom kernels, GPU profiling, compiler optimization, or high-performance computing is a plus.
Experience contributing to initiatives such as NanoGPT Speedrun, Marin, or similar open-source model-training projects is a strong plus.
Advanced degrees are welcome but not required.

Responsibilities

Research, design, and implement new techniques for training and improving large language models.
Build and optimize large-scale pre-training and post-training pipelines.
Improve model training efficiency, throughput, stability, and scalability.
Work on distributed training across large GPU clusters.
Design and optimize model-parallel training strategies, including tensor, pipeline, sequence, and data parallelism.
Optimize GPU workloads using technologies such as CUDA and Triton.
Improve inference and training kernels when necessary.
Explore new model architectures, training methodologies, and post-training techniques.
Run experiments, analyze results, and rapidly iterate on research ideas.
Collaborate on software/hardware co-design to maximize training throughput.
Contribute to internal research infrastructure and potentially open-source initiatives.