IBM Research Intern 2027: AI Systems

IBM; Cambridge (MA); On-site; USD 44,000 - 69,000; Full time

Job summary

IBM Research Cambridge invites a PhD student to participate in a 2027 summer internship within the AI Native Systems group to advance efficient, reliable agentic AI for IBM Z. The role centers on large language models, AI agents, adaptive memory management, and enterprise infrastructure in Finn/Paver frameworks.

You will collaborate with IBM researchers and academia, implement prototypes in Python, and contribute to publications and potential IP disclosures with real-world impact.

Qualifications

• PhD candidate in Computer Science, Electrical Engineering, or a related field.
• Experience with machine learning, LLMs, AI agents, NLP.
• Strong Python programming and ML framework experience (PyTorch, TensorFlow).

Responsibilities

• Research, design, and prototype adaptive information-control mechanisms for agentic AI systems.
• Define and model minimal sufficient agent state including observations, memory, and task progress.
• Develop and evaluate information acquisition, retrieval, summarization, and memory management techniques.
• Implement experimental solutions in Python and integrate into Finn/Paver or similar agent framework.
• Design experiments on z/OS management tasks measuring reliability, memory, latency, and efficiency.
• Analyze tradeoffs between performance, information efficiency, cost, and on-platform execution.
• Collaborate with researchers to review results and translate findings into practical architectures.
• Document outcomes through reports, presentations, demos, or IP disclosures.

Skills

Python

PyTorch

TensorFlow

NLP

AI research

Large language models

Education

Doctorate Degree

Tools

Hugging Face

Job description

Introduction

IBM Research is seeking a highly motivated PhD student to join the AI Native Systems organization for a 2027 summer internship focused on advancing the efficiency, reliability, and scalability of agentic AI systems for IBM Z. The selected candidate will conduct research at the intersection of large language models, AI agents, machine learning systems, adaptive memory management, and enterprise infrastructure, developing novel techniques that enable agents to acquire, retain, retrieve, summarize, and utilize only the information necessary to successfully complete complex, long-horizon system-management tasks. Working closely with IBM Research scientists and academic collaborators, the intern will design, implement, and evaluate adaptive information-control mechanisms within the Finn/Paver agent framework, contributing to next-generation AI capabilities for IBM Z and Spyre-enabled environments while gaining experience in cutting-edge research with potential impact on future products, publications, and intellectual property.

Your Role And Responsibilities

• Research, design, and prototype adaptive information-control mechanisms for agentic AI systems operating in IBM Z environments.
• Define and model minimal sufficient agent state, including observations, retrieved evidence, memory, interaction history, tool outputs, and task progress information.
• Develop and evaluate techniques for information acquisition, retrieval, summarization, memory management, context compression, and evidence-sufficiency estimation in long-horizon AI agent workflows.
• Implement experimental solutions in Python and integrate selected approaches into the Finn/Paver system-management agent or representative agent frameworks.
• Design and execute experiments on representative z/OS management tasks, measuring task completion reliability, context utilization, memory consumption, tool usage, latency, and inference efficiency.
• Analyze tradeoffs between agent performance, information efficiency, computational cost, and system resource requirements for on-platform execution using IBM Spyre and future IBM Z AI accelerators.
• Collaborate with IBM Research scientists and academic partners to review results, refine algorithms, and translate research findings into practical agent architectures.
• Document research outcomes through technical reports, presentations, demonstrations, and potential publications or intellectual property disclosures.

Preferred Education

Doctorate Degree

Required Technical And Professional Expertise

• Currently pursuing a PhD in Computer Science, Computer Engineering, Electrical Engineering, Artificial Intelligence, Machine Learning, or a related technical field.
• Demonstrated experience in machine learning, deep learning, large language models (LLMs), AI agents, natural language processing, or related AI research areas.
• Strong programming skills in Python and experience with machine learning frameworks such as PyTorch, TensorFlow, or equivalent.
• Experience designing and evaluating experiments, analyzing results, and developing research prototypes in AI, machine learning, or distributed systems.
• Excellent written and verbal communication skills, with the ability to document technical work and present research findings to technical and business audiences.

Preferred Technical And Professional Experience

• Research experience in agentic AI, large language models (LLMs), retrieval-augmented generation (RAG), AI memory systems, or long-context reasoning.
• Experience developing, fine-tuning, or evaluating AI/ML models using PyTorch, Hugging Face, or related frameworks.
• Knowledge of AI systems topics such as context compression, information retrieval, memory management, planning, tool use, or inference optimization.
• Experience conducting independent research resulting in publications, open-source contributions, patents, technical reports, or academic projects.
• Familiarity with Linux-based environments, distributed systems, cloud infrastructure, enterprise computing platforms, or system administration workflows.
• Interest in AI efficiency, privacy, security, and trustworthy AI systems.