Machine Learning Systems Engineer

As a Machine Learning Engineer, you will contribute directly to our machine learning infrastructure, to the ScalarLM open source codebase, and build large-scale language model applications on top of it. You'll operate at the intersection of high-performance computing, distributed systems, and cutting-edge machine learning research, developing the fundamental infrastructure that enables researchers and organizations worldwide to train and deploy large language models at scale.

This is an opportunity to take on technically demanding projects, contribute to foundational systems, and help shape the next generation of intelligent computing. You Will:

• Contribute code and performance improvements to the open source project.
• Develop and optimize distributed training algorithms for large language models.
• Implement high-performance inference engines and optimization techniques.
• Work on integration between vLLM, Megatron-LM, and HuggingFace ecosystems.
• Build tools for seamless model training, fine-tuning, and deployment.
• Optimize performance of advanced GPU architectures.
• Collaborate with the open source community on feature development and bug fixes.
• Research and implement new techniques for self-improving AI agents.

Experience Level: 3+ years of experience in machine learning engineering or research, * Programming Languages: Proficiency in both C/C++ and Python

• High Performance Computing: Deep understanding of HPC concepts, including:
• MPI (Message Passing Interface) programming and optimization
• Bulk Synchronous Parallel (BSP) computing models
• Multi-GPU and multi-node distributed computing
• CUDA/ROCm programming experience preferred
• Machine Learning Foundations:
• Solid understanding of gradient descent and backpropagation algorithms
• Experience with transformer architectures and the ability to explain their mechanics
• Knowledge of deep learning training and its applications
• Understanding of distributed training techniques (data parallelism, model parallelism, pipeline parallelism, large batch training, optimization)

Research and Development

• Publications: Experience with machine learning research and publications preferred
• Research Skills: Ability to read, understand, and implement techniques from recent ML research papers
• Open Source: Demonstrated commitment to open source development and community collaboration

Experience

• 3+ years of experience in machine learning engineering or research.
• Experience with large-scale distributed training frameworks (Megatron-LM, DeepSpeed, FairScale, etc.).
• Familiarity with inference optimization frameworks (vLLM, TensorRT, etc.).
• Experience with containerization (Docker, Kubernetes) and cluster management.
• Background in systems programming and performance optimization.
• PhD or MS in Computer Science, Computer Engineering, Machine Learning, or related field.
• Experience with SLURM, Kubernetes, or other cluster orchestration systems.
• Knowledge of mixed precision training, data parallel training, and scaling laws.
• Experience with transformer architecture, pytorch, decoding algorithms.
• Familiarity with high performance GPU programming ecosystem.
• Previous contributions to major open source ML projects.
• Experience with MLOps and model deployment at scale.
• Understanding of modern attention mechanisms (multi-head attention, grouped query attention, etc.).

At RelationalAI, you will:

• Work from anywhere in the world
• Earn competitive salary + equity
• Enjoy open PTO, flexible schedules, and recharge weeks
• Access global benefits, mental-health support, and learning stipends
• Join a transparent, inclusive, and globally connected culture that values curiosity, excellence, and impact
• Regular team offsites and global events - Building strong connections while working remotely through team offsites and global events that bring everyone together.
• A culture of transparency & knowledge-sharing - Open communication through team standups, fireside chats, and open meetings.

Country Hiring Guidelines:

RelationalAI hires people from around the world. All of our roles are remote; however, some locations might carry specific eligibility requirements.

Because of this, understanding location & visa support helps us better prepare to onboard our colleagues.

At RelationalAI, we're solving one of the most important challenges in artificial intelligence: how to teach large language models the logic, semantics, and business context of the modern enterprise. Frontier models are trained almost entirely on public data - they can speak about the world, but they don't understand your business. We fix that. RelationalAI has pioneered a breakthrough called Superalignment - technology that enables LLMs to learn natively from private, structured enterprise data inside the data cloud. By combining this with relational knowledge graphs and our proprietary neuro/symbolic-relational reasoners, we deliver trustworthy decision intelligence: systems that use semantic models to truly understand how a business operates and can reason across its data to drive better outcomes. We're a globally distributed team of engineers, scientists, and builders redefining how AI learns from data. We believe that high-stakes decisions deserve frontier intelligence - intelligence that's explainable, aligned, and grounded in reality. If you're driven by curiosity, thrive in complexity, and want to help build the system that brings true understanding to enterprise AI, you'll feel right at home here., ScalarLM unifies vLLM, Megatron-LM, and HuggingFace for fast LLM training, inference, and self-improving agents-all via an OpenAI-compatible interface. ScalarLM builds on top of the vLLM inference engine, the Megatron-LM training framework, and the HuggingFace model hub. It unifies the capabilities of these tools into a single platform, enabling users to easily perform LLM inference and training, and build higher lever applications such as Agents with a twist - they can teach themselves new abilities via back propagation. ScalarLM is inspired by the work of Seymour Roger Cray (September 28, 1925 - October 5, 1996), an American electrical engineer and supercomputer architect who designed a series of computers that were the fastest in the world for decades, and founded Cray Research, which built many of these machines. Called "the father of supercomputing", Cray has been credited with creating the supercomputer industry. It is a fully open source project (CC-0 Licensed) focused on democratizing access to cutting-edge LLM infrastructure that combines training and inference in a unified platform, enabling the development of self-improving AI agents similar to DeepSeek R1. ScalarLM is supported and maintained by TensorWave in addition to RelationalAI.