Machine Learning Engineer - New Grad

As a new grad Machine Learning Engineer, you’ll train and evaluate models that make large language models cheaper and more capable to serve. Depending on your interests, that could mean training speculator (draft) models and tuning acceptance rates, compressing models without losing accuracy, or designing and evaluating agentic workflows for tool use, planning, and long-horizon tasks. You’ll work closely with the AI performance, compiler, and runtime teams to turn research results into measured end-to-end wins, and you’ll have room to pursue open problems as our reinforcement learning and post-training work ramps up., * Train, evaluate, and tune speculative decoding models (Dspark/Dflash, Medusa/Eagle-style heads, and multi-token prediction) to increase acceptance rates on production workloads.

• Develop and implement optimization techniques, including quantization, sparsity, distillation, pruning, and parameter-efficient fine-tuning, to quantify the trade-offs between model quality, latency, and resource efficiency.
• Build and evaluate agentic systems: tool calling, multi-step planning, retrieval, and long-context reasoning on top of our serving stack.
• Design rigorous evaluations and benchmarks; make accuracy and quality claims that hold up under scrutiny
• Translate recent research into working implementations, and run ablations to understand what actually drives the result
• Partner with the AI performance, compiler, runtime, and product teams to ship model improvements into SambaNova products
• Contribute to emerging areas such as reinforcement learning, post-training, and reward modeling as the roadmap expands

Machine Learning, Artificial Intelligence (AI), GPU (Graphics Processing Unit), Energy Efficiency, Performance Modeling, Machine Learning, Reinforcement Learning, Proof of Concept, Product Shipments, Scientific Research, * B.S. in Computer Science, Electrical Engineering, Mathematics, or a related field

• Strong programming skills in Python
• Solid foundation in machine learning, linear algebra, probability, and numerical computing
• Hands-on experience with PyTorch, TensorFlow, or JAX
• Working knowledge of transformer architectures and the LLM training or inference pipeline
• Ability to design a clean experiment, interpret results honestly, and communicate findings clearly

Preferred Qualifications

• M.S. or Ph.D. in Computer Science, Electrical Engineering, or a related field
• Research or internship experience in LLM training, post-training, or inference
• Experience with speculative decoding, distillation, quantization, sparsity, or other model efficiency techniques
• Experience building agentic systems, tool-calling pipelines, or LLM evaluation harnesses
• Familiarity with reinforcement learning or preference optimization (RLHF, DPO, GRPO, reward modeling)
• Experience with distributed training frameworks (FSDP, DeepSpeed, Megatron) or inference frameworks such as vLLM, SGLang, or TensorRT-LLM
• Publications at ML venues (NeurIPS, ICML, ICLR, ACL, EMNLP) or meaningful open source contributions

SambaNova offers a competitive total rewards package, including the base salary, plus equity and benefits. We cover 95% premium coverage for employee medical insurance, and 77% premium coverage for dependents and offer a Health Savings Account (HSA) with employer contribution. We also offer Dental, Vision, Short/Long term Disability, Basic Life, Voluntary Life, and AD&D insurance plans in addition to Flexible Spending Account (FSA) options like Health Care, Limited Purpose, and Dependent Care. Our library of well-being benefits available to you and your dependents includes a full subscription to Headspace, Gympass+ membership with access to physical gyms, One Medical membership, counseling services with an Employee Assistance Program, and much more.

SambaNova is a leader in next-generation AI infrastructure, delivering a full-stack inference platform for customers worldwide. At the core of SambaNova’’s technology is the RDU (Reconfigurable Dataflow Unit) - a chip built on a dataflow architecture rather than the traditional GPU model. Its decode performance is especially strong for agentic workloads like multi-turn agents, code generation, and long-running applications. RDUs are packaged into SambaRack, rack-scale hardware that lets customers deploy state-of-the-art models with better performance, greater energy efficiency, and faster time to value.