Applied AI/ML Scientist

Role Overview

Applied AI/ML Scientist at Cerebras Systems. The role is based at Cerebras Systems, which builds the world's largest AI chip—56 times larger than GPUs—delivering over 10 times faster inference than GPU-based hyperscale cloud services.

Company Overview

Cerebras Systems develops the Wafer-Scale Engine (WSE), an architecture that enables industry-leading training and inference speeds for AI applications. The company partners with leading model labs, global enterprises, and AI-native startups, including a multi-year partnership with OpenAI to deploy 750 megawatts of scale for ultra high-speed inference.

Role Purpose

As an Applied AI Scientist in the FieldML team, you will develop and customize large language models and large-scale deep learning models to solve specific customer problems. You will bridge state-of-the-art research and real-world applications, helping customers harness the Cerebras Wafer-Scale Engine for their AI initiatives. You will implement, train, and scale models to solve complex business and scientific problems across diverse projects.

Key Responsibilities

Customer Use Case Discovery & Project Scoping

• Collaborate with customer stakeholders to identify the best approaches to their business problems with AI.
• Contribute to technical scoping of engagements, including feasibility analysis, data quality and availability assessments, and selection of optimal model architectures.
• Define project milestones, success metrics, and rigorous evaluation benchmarks to ensure the solution delivers measurable value to the customer's business.

Custom Models and AI Systems Development

• Architect and execute end-to-end training recipes for custom models, tailoring model architecture and training recipes to meet customer-specific performance and accuracy requirements.
• Design and implement sophisticated adaptation strategies, including continuous pre-training on private datasets, supervised fine-tuning (SFT), and post-training alignment via RLHF or DPO.
• Own the training pipeline from high-performance data preprocessing and tokenization through hyperparameter tuning and loss-curve analysis.
• Navigate model convergence on specialized hardware, performing deep-dive analysis into loss dynamics and gradient stability.
• Scale training workloads across Cerebras clusters, ensuring efficient utilization of hardware for multi-billion parameter models.
• Build and optimize core components of agentic systems, focusing on tool-use capabilities, long-context reasoning, and multi-step planning.

Technical Customer Leadership

• Serve as an AI/ML subject matter expert during technical deep-dives, translating customer requirements into precise training recipes.
• Build and maintain strong customer relationships to become their go-to AI/ML expert.

Internal Research and Engineering Collaboration

• Act as the voice of the customer for internal R&D and engineering teams to drive improvements in the software stack and hardware utilization.
• Partner with internal ML teams and product teams on prioritization of novel model architectures with the Cerebras software stack, development of training recipes, and internal case studies.
• Distill customer-facing successful projects into internal playbooks to scale the FieldML team's ability to deliver specialized models.

Qualifications & Experience

• Master's degree or PhD in Computer Science, Machine Learning, or related fields.
• Proven track record of training and/or fine-tuning large models with 1 billion or more parameters.
• Direct experience with the challenges of large-scale model training.

Skills & Competencies

• Expert-level understanding of modern model architectures, including dense transformers, MoEs, multimodal and sequence models, scaling laws, and training dynamics.
• Mastery of Python and PyTorch.
• Experience with distributed training frameworks and large-scale distributed data processing pipelines and tools.
• Strong interpersonal and communication skills, effective in collaborative and fast-paced team settings.
• Ability to work autonomously and within a team in a dynamic environment, managing multiple projects and pivoting as customer needs evolve.
• Ability to present complex technical results to diverse audiences—from C-level executives to research scientists.

Additional Information

Cerebras Systems is committed to creating an equal and diverse environment and is an equal opportunity employer. The company celebrates different backgrounds, perspectives, and skills and believes inclusive teams build better products and companies.