Automation Architect

About the Role

We are hiring an Agentic AI Automation Architect to shape how our clients — and we ourselves — use agentic AI to automate meaningful work. This is an architect role, not an AI engineer role: you will spend a significant part of your time with clients, framing business problems, identifying where agentic automation actually pays off, and designing solutions that are safe, observable, and scalable. You will also own the internal accelerators that make our delivery teams faster on every subsequent engagement.

Our current delivery stack anchors on the Microsoft agentic ecosystem — Copilot Studio, Azure AI Foundry, Semantic Kernel, Azure OpenAI, and the broader Azure AI services — integrated with Microsoft 365, Fabric, Dynamics, and enterprise systems via MCP and native connectors. You should be fluent enough in this stack to make credible architecture decisions. But the role is stack-savvy, not stack-loyal. We are looking for architects who have designed multi-agent solutions across tools over their career, and who can evaluate trade-offs between orchestration frameworks, memory stores, tool/MCP integrations, and model choices rather than defaulting to one.

The craft we are hiring for is a clear progression from business process to agent design to production-grade automation: starting with a precise workflow and its pain points, choosing the right agent topology (single agent, multi-agent, human-in-the-loop), designing tool use and memory thoughtfully, and engineering the guardrails, evaluation, and observability needed to run it in a regulated enterprise.

A candidate whose CV is primarily a collection of prompts, fine-tunes, or single-framework demos however impressive, is likely a strong AI engineer, not the architect we are hiring.

Key Responsibilities

Client Engagement & Opportunity Shaping

• Engage directly with client stakeholders Engage directly with client business and technology stakeholders to understand their processes, pain points, and where agentic automation genuinely fits (and where it does not).
• Translate vague AI ambitions Translate vague AI ambitions into a prioritised portfolio of agentic use cases with clear success metrics, ROI hypotheses, and risk profiles.
• Run discovery workshops Run discovery workshops, present agentic architectures, and defend design choices in front of both technical and non-technical audiences — including risk, legal, and compliance functions.
• Act as the trusted advisor Act as the trusted advisor across the engagement lifecycle — pre-sales, solutioning, delivery oversight, and post-go-live scaling.

Agentic Solution Architecture

• Own the target-state architecture Own the target-state architecture for agentic solutions: agent topology (single agent, multi-agent, supervisor / worker, human-in-the-loop), orchestration model, and integration pattern into enterprise systems.
• Design tool use rigorously Design tool use rigorously — which tools, when to use them, how they are described, and how they are exposed (including MCP servers and native connectors).
• Design memory and state Design memory and state — short-term context, long-term memory stores, retrieval strategy, and what the agent should and should not remember across sessions.
• Choose the right model for each task Choose the right model for each task (reasoning vs. fast, proprietary vs. open, grounded vs. creative) and justify the choice in cost, latency, and quality terms.
• Define reference architectures Define reference architectures and architecture decision records (ADRs); establish patterns that multiple delivery teams can reuse.

Delivery on the Microsoft Agentic Stack

• Shape solutions across the Microsoft agentic stack Shape solutions across Copilot Studio (low-code agents and Copilot extensions), Azure AI Foundry (pro-code agent development, model catalogue, evaluations), Semantic Kernel, Azure OpenAI, and Azure AI Search — at a design and oversight level.
• Integrate agents with enterprise systems Integrate agents with Microsoft 365 (Graph, Teams, Outlook), Dynamics 365, Microsoft Fabric, and external enterprise systems via MCP, custom connectors, and Logic Apps / Power Automate where appropriate.
• Guide the delivery team Guide the delivery team on environment setup, model provisioning, capacity and cost management, and deployment patterns. You will review and direct, not build every agent yourself.

Safety, Governance & Evaluation

• Architect responsible AI guardrails Architect responsible AI guardrails: content filters, prompt injection defences, PII handling, tool-use authorisation, and human-in-the-loop checkpoints for high-risk actions.
• Define evaluation strategies Define evaluation strategies — offline evals, LLM-as-judge, red-teaming, and production telemetry — so that agent quality is measurable before and after release.
• Design observability Design observability: tracing, conversation logs, tool-call audit trails, cost and token monitoring, and feedback loops.
• Ensure compliance Ensure compliance with relevant standards (GDPR, DPDP, HIPAA, EU AI Act, or industry-specific) as part of every design.

Internal Accelerators & Capability Building

• Own the internal agentic automation roadmap Own the internal agentic automation roadmap — identify where our own delivery, pre-sales, and operations can be automated, and lead the build of reusable accelerators.
• Build a library of reference agents Build a library of reference agents, tool/MCP integrations, guardrail patterns, and evaluation harnesses that every engagement starts from.
• Mentor AI engineers, automation developers, and junior architects Mentor AI engineers, automation developers, and junior architects; raise the design quality of the broader practice.
• Support pre-sales Support pre-sales — RFP responses, solution estimates, and PoCs — as part of the Capability function.

Required Qualifications

• Up to 14 years of total experience Up to 14 years of total experience, with the last several years spent in an architecture or lead-solutioning capacity across AI, automation, or intelligent applications — not purely in hands-on engineering.
• Demonstrable track record of end-to-end AI or automation solutions Demonstrable track record of designing end-to-end AI or automation solutions that went to production, ideally across more than one framework or stack over your career.
• Hands-on design experience with agentic frameworks Hands-on design experience with agentic frameworks — Copilot Studio, Azure AI Foundry, and Semantic Kernel at a minimum. Exposure to LangGraph, AutoGen, CrewAI, or similar is a plus; MCP literacy is expected.
• Strong grounding in LLM application patterns Strong grounding in LLM application patterns: RAG, tool use and function calling, structured outputs, prompt engineering, context engineering, and multi-agent orchestration.
• Working fluency in Python Working fluency in Python; comfortable reading C#, TypeScript, and SQL; you will review and challenge code, even if you are not the primary author on most engagements.
• Solid understanding of the Azure ecosystem Solid understanding of the Azure ecosystem — identity (Entra ID), networking, data services (including Microsoft Fabric and Azure AI Search), and cost management.
• Strong grounding in responsible AI Strong grounding in responsible AI, security, data governance, and compliance as applied to generative and agentic systems.
• Excellent client-facing communication Excellent client-facing communication — able to lead workshops, present to CXOs and risk functions, and produce crisp architecture documents in English.
• Willingness to work on-site Willingness to work on-site from our Gurgaon office and travel to client locations as required.

Good to Have

• Microsoft certifications: AI-102 (Azure AI Engineer), AI-900, or architecture-level certifications (Azure Solutions Architect Expert).
• Prior experience with RPA platforms (UiPath, Power Automate, Automation Anywhere) and a clear view of where classical RPA ends and agentic automation begins.
• Experience evaluating and integrating open-weight models, self-hosted inference, and fine-tuning / distillation where enterprise constraints require it.
• Domain depth in one of: BFSI, healthcare / life sciences, retail & CPG, or manufacturing — enough to recognise which workflows are worth automating and which are not.
• Pre-sales experience — responding to RFPs, building PoCs, and estimating AI / automation programmes.
• Contributions to open-source agent frameworks, MCP servers, or published writing on agentic architecture.

What We Value

We value architects who think in business outcomes first and technology second — and who are honest about where agents add value and where they do not. The best agentic solutions start with a precise process, rest on a thoughtful agent and tool design, and end with automation that decision-makers actually trust in production. If that is how you think about agentic AI — and you want to do it in front of clients, with a strong delivery team behind you — we would like to talk.