5 days ago
Bengaluru, IndiaMid Level

Responsibilities

  • Build and maintain LLM harnesses with agent loops, tool/function calling, context construction, memory, retries, and failure handling.
  • Design, iterate, version, and evaluate prompts using structured experiments and evaluation datasets.
  • Build and operate RAG pipelines covering document extraction, chunking, indexing, retrieval, reranking, prompt assembly, and response handling.
  • Generate embeddings, manage vector indexes, and tune retrieval quality and cost.
  • Build document, structured-record, and customer-data parsing and curation pipelines.
  • Write accuracy-validation scripts, maintain golden and regression evaluation datasets, and report quality metrics.
  • Apply classification, clustering, lightweight fine-tuning, data analysis, sample inspection, and error analysis to support AI design decisions.
  • Deliver AI-side Strike Team commitments and production-quality tested code.

Requirements

  • Bachelor's or Master's degree in Computer Science, Data Science, Statistics, or a related technical field.
  • 3+ years of AI or data science experience, including hands-on LLM-based application development.
  • Strong Python skills with pandas, numpy, and scikit-learn.
  • Hands-on experience building LLM harnesses with agent loops, tool/function calling, and structured outputs against Anthropic, OpenAI, or Bedrock APIs.
  • Experience with prompt engineering, prompt versioning, prompt evaluation, and at least one orchestration framework and vector database.
  • Working knowledge of embeddings, similarity search, retrieval evaluation, and classical ML.
  • Comfort using Claude Code as an AI coding assistant.
  • Ability to write clean, tested production code and familiarity with Git, code review, and CI/CD.
  • Preferred experience includes LangGraph, Claude Agent SDK, OpenAI Agents, custom production agent harnesses, structured eval harnesses, AWS, Bedrock, SageMaker, OpenSearch, model fine-tuning or distillation, Guidewire products, insurance or finance datasets, and document-heavy or regulated domains.

Tech Stack

AWSGitNumPyPandasPythonscikit-learn

Categories

Guidewire Software

About Guidewire Software

1,001-5,000 employees

Guidewire is the platform P&C insurers trust to engage, innovate, and grow efficiently. More than 570 insurers in 43 countries, from new ventures to the largest and most complex in the world, rely on Guidewire products. With core systems leveraging data and analytics, digital, and artificial intelligence, Guidewire defines cloud platform excellence for P&C insurers. We are proud of our unparalleled implementation record, with 1,700+ successful projects supported by the industry’s largest R&D team and SI partner ecosystem. Our marketplace represents the largest partner community in P&C, where customers can access hundreds of applications to accelerate integration, localization, and innovation. For more information, please visit https://www.guidewire.com/.

Contact me