About the role
## About the role You architect products that put large language models and agents into production for clients: deciding where AI belongs and where it does not, designing the agent, retrieval and evaluation layers, and building the application architecture around them on Azure. You own the architecture from the first design decision through production and lead the engineers who build it. You will work across engagements and domains. **Applied AI experience is the deciding requirement for this role. Azure platform or infrastructure architecture experience alone does not qualify.** ## What you will do - Set the AI boundary on every product: what runs deterministically, what the model handles, and how human review and evaluation gate each release. - Design agent and retrieval layers: Azure OpenAI behind a swappable model port, Semantic Kernel or Microsoft.Extensions.AI (or the Node equivalents) for orchestration and tool calling, Azure AI Search for retrieval-augmented generation, and Azure AI Document Intelligence for document ingestion. - Define how AI features are measured: evaluation sets, error thresholds, regression runs in CI, cost and latency telemetry, and the human-review workflow that catches what evaluations miss. - Design the application architecture around the AI: hexagonal systems with a generic domain core and client-specific adapter packs, enforced by contract tests and lint rules. - Design effective-dated and bitemporal data models on Azure SQL temporal tables, and reliable integration through the transactional outbox pattern over Azure Service Bus. - Plan legacy modernization as strangler-fig migrations: a facade over the legacy system, feature-flagged rollout, parity checks, and an observable checkpoint per stage. - Define embedded UI and identity contracts: web components consuming a host platform's design tokens, and federation over OpenID Connect or SAML. - Own C4 diagrams and architecture decision records, run design and code reviews, mentor the team, and present decisions to client leadership with costs and failure modes stated plainly. ## Required - LLM applications shipped to production: tool-calling agents, structured outputs, retrieval-augmented generation, and prompt and model versioning, on Azure OpenAI or an equivalent provider. - Evaluation and guardrails in production: golden sets, error taxonomies, regression on prompt changes, human-review sampling, content filtering. You can describe a release an evaluation blocked and why. - Judgment about AI placement: you can explain, with examples from your own work, where a model should not be used and what deterministic design replaced it. - 13+ years in software engineering, including 5+ years as a software or solutions architect on systems in production. - Hands-on Azure application architecture in .NET (C#) or Node.js (TypeScript): modular monoliths, Azure SQL, Azure Service Bus, container or app hosting, CI/CD, infrastructure as code. - Domain-driven designs shipped with hexagonal or clean architecture, including an anti-corruption layer against a legacy system. - Temporal or effective-dated data models and reliable messaging (outbox, idempotency, replay) in production. - Clear architecture communication: C4 or equivalent diagrams, decision records, and the ability to brief a CTO and a non-technical product leader in the same meeting. ## Preferred - Multi-tenant SaaS design: tenant isolation, per-tenant configuration, single-tenant to pooled tenancy without a rewrite. - Legacy modernization delivered with strangler-fig migration, feature flags, and parity validation. - Depth in both .NET and Node.js; Python for AI services. - Azure AI Document Intelligence and document generation or rendering pipelines. - Third-party integrations behind provider abstractions: e-signature, payments, messaging, data feeds. - Web components, design-token systems, or micro-frontend composition; OpenID Connect or SAML federation. - Cost modeling for AI workloads: token spend, provider fees, run cost per tenant. - Azure Solutions Architect Expert or Azure AI Engineer certification. ## Technology environment - **AI:** Azure OpenAI, Azure AI Search, Azure AI Document Intelligence, Azure AI Foundry for evaluations and prompt management, Azure AI Content Safety; Semantic Kernel and Microsoft.Extensions.AI, or Vercel AI SDK, LangChain.js and the OpenAI SDK. - **Backend:** .NET 8 (C#) or Node.js (TypeScript), modular monolith, domain-driven design. - **Cloud:** Azure App Service or Container Apps, Azure SQL (temporal tables), Azure Service Bus, Azure Functions, Key Vault, Azure DevOps, infrastructure as code (Bicep or Terraform), Application Insights and OpenTelemetry. - **Frontend:** TypeScript, web components, design-token systems. - **Patterns:** hexagonal architecture, anti-corruption layers, transactional outbox, bitemporal data, strangler fig, feature flags, single-tenant deployment stamps designed to pool later. - **Documents and integrations:** OpenXML or docx libraries, server-side PDF rendering, e-signature and other third-party providers behind abstractions. ## Engagement details - Client requisition title: Lead AIML Engineer - Number of openings: 1 - Work location: Remote (India). **Candidates based in Hyderabad are not eligible for this role** — the client has excluded that location. - Employment type: Full-time - Work timing / shift: 11 AM – 8 PM IST - Notice period: immediate joiners preferred - Relevant experience: 5+ years as an architect - Total experience: 13+ years
Skills: Azure OpenAI, LLM Agents, RAG, Solution Architecture, LLM Evaluation, Azure AI Search, Semantic Kernel, .NET 8, C#, Node.js, TypeScript, Domain-Driven Design, Hexagonal Architecture, Azure SQL, Azure Service Bus, Transactional Outbox, Strangler Fig, Bicep, Terraform, Azure DevOps