Agentic AI — PhD Intern; Autonomous AI Agent

Position: Agentic AI — PhD Intern (Autonomous AI Agent)
Overview About the Program

Keysight's AI team builds adaptive engineering intelligence systems where measurement, simulation, and machine learning converge - models that learn from complex physical data and adapt to new design and testing conditions.

This internship focuses on reliable autonomous AI agent systems for multi-step engineering workflows. You will build and evaluate an AI workflow that can maintain context across longer-running tasks, use approved tools, incorporate new information, and help engineers complete work more efficiently.

Role Overview

You will design, build, and evaluate an autonomous agentic workflow for a defined engineering process. The system will interpret incoming information, determine appropriate next steps, use approved tools to complete tasks, track progress, and incorporate feedback as the workflow evolves.

A central constraint is that the system's actions must be bounded, auditable, and recoverable. Engineers need to understand what it did, why it acted, what remains open, and where approval is required. You will develop evaluations for task completion, continuity, reliability, and human intervention. You will work alongside AI researchers, software engineers, and engineering stakeholders for the selected workflows.

Learning outcomes include applying agentic AI to realistic multi-step engineering workflows; designing stateful AI systems with durable context, provenance, recovery, and human approval points; evaluating continuity, reliability, completion rate, and intervention rate for non-deterministic systems; and building fluency with tool integration, workflow automation, task state management, and auditability requirements.

Responsibilities

• Build an autonomous agentic workflow that maintains task state and context across multiple steps and sessions.
• Develop tool integrations connecting the system to approved communication, planning, document, workflow, and engineering applications.
• Implement mechanisms for incorporating new information, updating task state, and resuming incomplete work.
• Develop appropriate state and provenance mechanisms for decisions, actions, outcomes, and feedback.
• Design guardrails for approval, permissions, bounded actions, failure recovery, and escalation.
• Assemble representative engineering workflows and measure completion rate, reliability, continuity, and human intervention against a suitable baseline.

Required Qualifications

• Currently pursuing or recently completed a PhD in Machine Learning, Computer Science, Electrical or Electronic Engineering, Applied Mathematics, or a related technical field, with experience building and evaluating agent systems.
• Strong Python and software engineering: modular code, version control, reproducible pipelines.
• Hands-on experience building LLM agents - tool and function calling, orchestration, context management - well beyond prompt-level experimentation.
• Demonstrated rigor in evaluating non-deterministic systems: benchmark design, held-out test sets, automated scoring, and a healthy distrust of demonstrations.
• A portfolio of agent or LLM systems you have built: code we can look at, not descriptions, with an account of how you evaluated them.
• Working knowledge of application state, task management, context handling, and tool integration.
• Sound judgment about autonomy - where an agent should act, where it should ask, and where it should refuse.
• Clear communication to both AI researchers and domain engineers.

Preferred Qualifications

• Industry experience shipping agent or LLM-based systems to real users.
• Experience with agent frameworks, MCP or comparable tool-integration protocols, and stateful application design.
• Background in workflow automation, task orchestration, enterprise software, or event-driven…