About the Role:

Would you like to participate in creating the fastest Generative Models inference in the world? Join the Cerebras Inference Team to participate in development of unique Software and Hardware combination that sports best inference characteristics in the market while running largest models available. 

Cerebras wafer scale inference platform allows running Generative models with unprecedented speed thanks to unique hardware architecture that provides fastest access to local memory, ultra-fast interconnect and huge amount of available compute. 

You will be part of the team that works with latest open and closed generative AI models to optimize for the Cerebras inference platform. Your responsibilities will include working on model representation, optimization and compilation stack to produce the best results on Cerebras current and future platforms.  

Responsibilities:

• Analysis of new models from generative AI field and understanding of impacts on compilation stack
• Develop and maintain model definition framework that consists of model building blocks to represent large language models based on PyTorch and Cerebras dialects ready to be deployed on Cerebras hardware.
• Develop and maintain the frontend compiler infrastructure that ingests PyTorch models and produces an intermediate representation (IR).
• Extend and optimize PyTorch FX / TorchScript / TorchDynamo-based tooling for graph capture, transformation, and analysis.
• Collaboration with other teams throughout feature implementation
• Research on new methods for model optimization to improve Cerebras inference

 Qualifications:

• Degree in Engineering, Computer Science, or equivalent in experience and evidence of exceptional ability
• Strong Python programming skills and in-depth experience with PyTorch internals (e.g., TorchScript, FX, or Dynamo).
• Solid understanding of computational graphs, tensor operations, and model tracing.
• Experience building or extending compilers, interpreters, or ML graph optimization frameworks.
• Experience working with PyTorch and HuggingFace Transformers library
• Knowledge and experience working with Large Language Models (understanding Transformer architecture variations, generation cycle, etc.)
• Strong C++ programming skills.
• Knowledge of MLIR based compilation stack

Preferred Qualifications

• Prior experience contributing to PyTorch, TensorFlow XLA, TVM, ONNX RT, or similar compiler stacks.
• Knowledge of hardware accelerators, quantization, or runtime scheduling.
• Experience with multi-target inference compilation (e.g., CPU, GPU, custom ASICs).
• Understanding of numerical precision trade-offs and operator lowering.
• Contributions to open-source ML compiler projects.

About The Role

As an Applied AI Scientist in the FieldML team, you will be responsible for developing and customizing large language models and more broadly large-scale deep learning models to solve specific customer problems. You won't just advise; you will build. You will bridge the gap between state-of-the-art research and real-world applications by helping customers harness the power of the Cerebras Wafer-Scale Engine (WSE) for their AI initiatives.  

We are looking for experienced AI Scientists who are passionate about the "applied" side of machine learning - those who enjoy not just reading papers, but implementing, training, and scaling models to solve complex business and scientific problems. You will work on a diverse range of projects, from training bespoke models from scratch to fine-tuning and optimizing the latest Large Language Models (LLMs) for specific industry verticals, to designing and building components for custom agentic systems. 

The ideal candidate has experience in large model training and/or post-training, a deep understanding of training dynamics and model convergence, and expertise in data curation, combined with strong communication skills.   

Key Responsibilities 

• Customer Use Case Discovery & Project Scoping 
  • Collaborate with customer stakeholders to identify the best approaches to their business problem with AI. 
  • Contribute to the technical scoping of engagements, including feasibility analysis, data quality/availability/readiness assessments, and the 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 SOTA 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. 
  • Take full ownership of the training pipeline, from high-performance data preprocessing and tokenization to hyperparameter tuning and loss-curve analysis. 
  • Navigate the nuances of 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 the hardware for multi-billion parameter models. 
  • Build and optimize the 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 our software stack and hardware utilization. 
  • Partner with internal ML teams and product teams on prioritization of novel model architectures with Cerebras software stack, development of training recipes and internal case studies. 
  • Distill customer-facing successful projects into internal playbooks, helping scale the FieldML team’s ability to deliver specialized models.

Skills And Qualifications   

• Education: Master’s or PhD in Computer Science, Machine Learning, or related fields. 
• Broad Deep Learning Expertise: Expert-level understanding of modern model architectures, including dense transformers, MoEs, multimodal and sequence models, scaling laws and training dynamics.  
• Hands-on Trainig Experience: Proven track record of training and/or fine-tuning large models (1B+ parameters) and direct experience with the challenges of large-scale model training.  
• Engineering Proficiency: 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, able to work autonomously and within a team in a dynamic environment, managing multiple projects and pivoting as customer needs evolve. Able to present complex technical results to diverse audience - from C-level executives to research scientists, and to work collaboratively to solve customers’ unique challenges.  

• Develop design specifications for new machine learning and linear algebra kernels and mapping to the Cerebras WSE System using various parallel programming algorithms.
• Develop and debug kernel library of highly optimized low level assembly instruction and C-like domain specific language routines to implement algorithms targeting the Cerebras hardware system.
• Develop and debug high-performance kernel routines in low-level assembly and a custom C-like (CSL) language, implementing algorithms optimized for the Cerebras hardware system.
• Using mathematical models and analysis to measure the software performance and inform design decisions.
• Develop and integrate unit and system testing methodologies to verify correct functionality and performance of kernel libraries.
• Study emerging trends in Machine Learning applications and help evolve Kernel library architecture to address computational challenges of the start-of-the-art Neural Networks.
• Interact with chip and system architects to optimize instruction sets, microarchitecture, and IO of next generation systems.

• Bachelor’s, Master’s, PhD or foreign equivalents in Computer Science, Computer Engineering, Mathematics, or related fields.
• Understanding of hardware architecture concepts — must be comfortable learning the details of a new hardware architecture.
• Skilled in C++ and Python programming languages.
• Good knowledge of library and/or API development best practices.
• Strong debugging skills and knowledge of debugging complex software stack.

• Experience in kernel development and/or testing.
• Familiarity with parallel algorithms and distributed memory systems.
• Experience in programming accelerators such as GPUs and FPGAs.
• Familiarity with Machine Learning neural networks and frameworks such as TensorFlow and PyTorch.
• Familiarity with HPC kernels and their optimization.