Principal - Architecture

This is a role for a builder architect a seasoned application architect who has upskilled into a hands on GenAI practitioner someone who still writes code has put agentic and GenAI based solutions into live environments and can translate AIled engineering strategy into executable delivery models for customer stakeholders across time zones

Key Responsibilities

1 GenAI Solution Architecture SDLC Acceleration

Define the target state architecture and adoption roadmap for AIled SDLC where and how AI agents participate in planning coding review testing and deployment phases

Extend agents upstream into engineering design spec driven development specs and design documents as the source of truth agents implement against AIassisted architecture documentation and agent reviewers in design and codereview gates

Architect endtoend GenAI solutions agentic developer workflows RAG pipelines LLMbased applications for insurance workloads and scale successful patterns from pilot to production

Lead architecture transformation accelerated by AI corelegacy modernization and new eventdriven architecture EDA implementations using agents for code comprehension documentation migration functionalequivalence test generation and event schemahandler scaffolding

Design AIenabled business workflows with humans in the loop document intake and extraction confidence thresholds and exception routing reviewer feedback loops and endtoend auditability

Enable and coach offshore developer squads on agentassisted engineering define working practices quality gates and guardrails for AIgenerated code

Lead a team of GenAI engineers delivering these solutions setting technical direction reviewing designs and growing the teams agentic engineering capability

Measure and report acceleration outcomes cycle time throughput quality and adoption metrics and iterate the blueprint based on evidence

Define and enforce architecture standards design patterns and responsibleAI practices with particular attention to the data privacy security and regulatory constraints of the insurance industry operate architecture governance design authority decision records and reviews across delivery teams

Establish evaluation and guardrails as firstclass engineering evaluation harnesses and golden datasets hallucination and regression checks output guardrails and production monitoring for GenAI systems

2 Agentic Engineering Tooling handson

Design multiagent systems with defined roles orchestrator coder reviewer tester agents including task routing state management and humanintheloop checkpoints

Handson configuration and governance of GenAI developer tooling at enterprise scale agentic coding tools CLIdriven coding agents and AI pairprogramming solutions that go well beyond autocompletestyle copilots including toolserver MCP setup custom commands hooks and repository contextinstruction files plus agent SDKs and orchestration frameworks

Design and deploy MCP Model Context Protocol servers to expose the customers tools APIs and data sources to AI agents apply tooluse functioncalling patterns for LLMdriven agents

Select and integrate foundation models eg Anthropic Claude OpenAI GPT Gemini via APIs or managed platforms AWS Bedrock Azure OpenAI AI Foundry Google Vertex AI design modeltiering and routing strategies frontier reasoning models for complex work fast lightweight models for highvolume steps openweight models Llama Mistral where data residency requires balancing capability cost and latency

Deploy and operate GenAI workloads within enterprise cloud estates private endpoints and network isolation identity and access management quota ratelimit and regionalavailability planning model gateways and cost governance token budgeting caching chargeback

3 Context Engineering Business Ontology Domain Knowledge

Mine existing artifacts legacy code documentation wikis tickets to bootstrap the customers knowledge fabric establish curation versioning and freshness practices so context stays accurate as systems and regulations evolve

Data preparation for AI profile and cleanse source data normalization deduplication OCR noise in scanned documents structure unstructured content and curate versioned datasets including synthetic data where appropriate for retrieval evaluation and finetuning

Data protection in AI pipelines deidentification PII masking and redaction for data used in prompts retrieval corpora and evaluations data lineage and access controls aligned to insurance privacy and regulatory requirements

Metadata management and data governance business glossaries data catalogs and lineage that keep the ontology retrieval corpora and datasets trustworthy and traceable as systems evolve

Apply prompt engineering rigor few