Junior AI Engineer- Applied AI

Guidewire Software

3d

About The Position

Build, evaluate, and harden Guidewire Concierge, an omnichannel AI agent for P&C insurance carriers that handles policyholder calls, collects first notice of loss (FNOL) information, and creates claims in ClaimCenter. The product operates in production with real carriers and policyholders. The role focuses on agent orchestration, context, tool use, and evaluations that establish whether changes improve the system. Because the agent handles synchronous voice and regulated claims intake, latency and accuracy are important product requirements. The team increases the agent’s autonomy incrementally, using evidence to guide each step.

The engineer owns end-to-end portions of the agent, including prompt and context architecture, retrieval over policy and coverage data, integrations with core claims systems, and human handoff. The role also develops and operates evaluation processes, improves containment and resolution while preserving accuracy, instruments system performance, and contributes to the voice stack. It includes listening to carrier calls and recordings to identify conversation failures and opportunities to improve the product.

Requirements

• Have shipped an LLM application on which real users depend.
• Be fluent in LLMs, context engineering, retrieval-augmented generation (RAG), and agentic workflows, with underlying machine-learning and deep-learning depth.
• Use disciplined evaluations, error analysis, and statistical measurement to make unpredictable systems reliable enough to trust.
• Understand cost, latency, reliability, and scalability as engineering tradeoffs.
• Have designed data stores and distributed systems, write tests, take security and privacy seriously, and advocate CI/CD process automation.
• Work effectively with coding agents, including managing context, deciding when to write a specification, providing verifiers, and knowing when to intervene.
• Bring product sense and sufficient insurance context to shape product decisions and judge when to move quickly or slow down.
• Continue learning and adapting as AI tools and practices change.

Nice To Haves

• Experience with real-time or streaming voice systems, including ASR, TTS, telephony, or SBC/SIPREC.
• Experience in insurance, claims, or another regulated domain where errors have a cost.
• Experience building internal evaluation or observability tools adopted by other engineers.
• Knowledge of machine-learning principles, or experience experimenting with, training, and deploying models in production.
• Experience in P&C insurance or exposure to Guidewire products and services.
• Experience applying prompt or pipeline optimization frameworks such as DSPy or GEPA in production.

Responsibilities

• Own end-to-end parts of the agent, including prompt and context architecture, retrieval over policy and coverage data, tool and API calls into core claims systems, and human handoff.
• Build and run evaluation processes, including golden sets, trajectory evaluation, LLM-as-judge, and error analysis on production call traces; use evaluations to gate releases.
• Improve containment and resolution rates without sacrificing accuracy, and instrument the system to measure changes in both.
• Work on the voice path, including turn-taking, interruptions, latency budgets, and the telephony layer, such as LiveKit and SIP.
• Attend carrier calls and listen to recordings to identify where conversations fail and inform fixes.
• Help decide what the agent should do next, balancing MVP delivery with the need to get claims workflows right.

Benefits

• Health insurance
• Dental insurance
• Vision insurance
• Paid time off
• Company-sponsored retirement plan
• Some roles may be eligible for an annual company bonus plan, commissions, and/or long-term incentive awards, subject to applicable factors.

What This Job Offers

• Job Type: Full-time
• Career Level: Entry Level
• Education Level: No Education Listed