AI Architect

Roles & Responsibilities

Shape demand and executive conviction

Lead discovery and architecture workshops with client business and technology leadership; set the agenda rather than respond to it.

Identify pain points, value levers, and high-impact AI opportunities, and sequence them into a pragmatic AI roadmap (quick wins platform investments).

Translate ambiguous, complex client questions into solution hypotheses, target-state architecture, and a phased path to get there.

Build AI scenarios and value frameworks that illustrate meaningful business transformation, with measurable outcomes and a defensible business case.

Own the architecture

Own end-to-end architecture: LLM and model-portfolio selection, context engineering, retrieval and knowledge architecture, agent orchestration, microservices, integrations, data platform touchpoints, and UI.

Set the non-functional bar and design to it latency, throughput, unit cost per interaction, accuracy/eval thresholds, observability, safety, security, data residency, and auditability.

Define reference architectures, agentic blueprints, and accelerator standards for the practice; govern their reuse so pursuits start at 40% rather than zero.

Make and document build-vs-buy and platform decisions, including model routing, fine-tuning vs prompting vs retrieval, and open vs proprietary model strategy.

Prove it

Build prototypes, storyboards, demos, accelerators, and reusable agentic blueprints — personally on the critical path, and through the FDE pod for the rest.

Engineer and validate AI agents for reliability, reasoning, safety, and performance; establish the evaluation and red-teaming approach that lets us state quality claims with evidence.

Lead PoCs, pilots, and phase-0 MVPs across industries, arbitrating scope so the proof lands inside the sales cycle.

Win and hand over

Author and quality-gate client-ready content: architecture decks, SoWs, proposals, estimates, value frameworks, risk and assumption registers, and demo storylines.

Lead technical deep-dives, security and architecture reviews, and competitive differentiation discussions; handle the hardest objections in the room.

Refine solution narratives based on client feedback and market conditions.

Oversee a seamless transition from sales to delivery — architecture decision records, working code, known risks, and continuity of the people involved.

Build the practice

Create thought leadership and reusable IP; represent the firm externally at client forums, partner events, and in published material.

Mentor AI Leads and engineering teams; grow the bench of people who can do this role.

Collaborate with hyperscalers and product partners to craft next-generation AI-led experiences and co-funded initiatives.

Must-have skills and experience

Architecture and platform depth

Proven track record architecting and shipping production-grade GenAI/agentic systems in enterprise environments — not only pilots.

Deep command of LLM selection and evaluation, context engineering, model routing and cost control, and the trade-offs between prompting, retrieval, tool use, and fine-tuning.

Strong microservices and integration architecture: event-driven and API-led patterns, identity and access, secrets, multi-tenancy, and integration with core enterprise systems.

Enterprise concerns as first-class design inputs: security, privacy, data residency, model governance, responsible AI, and regulatory constraints.

Agentic AI engineering (hands-on)

Hands-on working knowledge of defining and configuring prompts, instructions, tools, reasoning strategies, guardrails, memory, and orchestration in AI agent development.

Design and debugging of multi-agent and long-horizon workflows, including failure modes, recovery, and human-in-the-loop control points.

Strong experience with RAG pipelines, vector databases, tokenization, and prompt engineering — including retrieval evaluation and knowledge-freshness strategy.

Engineering craft (hands-on)

Hands-on experience creating and maintaining Python libraries; working command of LangChain (or equivalent orchestration frameworks), Hugging Face, OpenAI/Anthropic APIs, and locally hosted open models.

Very strong experience developing RESTful APIs in Python with FastAPI (or similar) and integrating third-party services, UI components, and APIs.

Hands-on Docker-based deployment, CI/CD, and disciplined use of Git/GitHub for source control and review — mandatory.

Working knowledge of modern data platforms and their GenAI surfaces (Databricks, Snowflake Cortex, lakehouse patterns).

Model provider and hyperscaler GenAI platform know-how

Broad, current understanding of the GenAI offerings across the major frontier-model providers and hyperscalers, with deep, hands-on mastery of at least one — deep enough to design on it, defend it in a client architecture review, and know where it breaks:

Anthropic — Claude model family, Claude API, tool use and extended thinking, Claude Agent SDK, MCP (Model Context Protocol) as an integration standard, Claude delivered via Amazon Bedrock / Vertex AI, and enterprise deployment and governance patterns.

OpenAI — GPT model family, Responses/Assistants APIs, Agents SDK, function calling, structured outputs, fine-tuning, batch and realtime APIs, enterprise data-handling commitments.

AWS — Amazon Bedrock (multi-model access, Knowledge Bases, Guardrails, Agents/AgentCore), SageMaker, and integration with AWS data, identity and security services.

Google — Gemini model family, Vertex AI (Model Garden, grounding, agent tooling, evaluation), and Gemini enterprise offerings.

Microsoft Azure — Azure OpenAI / Azure AI Foundry, AI Agent Service, AI Search for retrieval, and Azure-native identity, networking and compliance patterns.

Able to construct and defend a multi-provider model strategy: capability-to-task mapping, model routing and tiering, benchmark and eval evidence, context-window and latency envelopes, unit economics, rate limits and capacity planning, deprecation and version-migration risk.

Fluent in the enterprise decision criteria that actually settle these choices — data residency and sovereignty, tenant isolation, IP and training-data commitments, certifications and auditability, regional availability, and the client's existing cloud and licensing commitments.

Designs for portability where it matters: provider-abstracted interfaces, gateway/routing layers, and prompt and eval assets that survive a model or vendor change, without over-abstracting to the point of losing provider-specific capability.

Keeps pace with a fast-moving field and translates new releases into concrete client opportunity, filtering hype from what is production-ready.

Consultative and commercial ability

Demonstrated ability to turn ambiguous business problems into crisp AI use cases with clear value hypotheses and measurable outcomes.

Excellent communication and articulation skills; credible with a CIO/CDO and with a client's principal engineer in the same day.

Commercial fluency: estimation, deal shaping, pricing implications of architecture choices, and the discipline to say no to a use case that will not survive contact with production.

Hands-on prototyping mindset — able to quickly build functional demos, agent workflows, or LLM-powered interactions that accelerate client conviction.

Good to have

Prior experience in a pre-sales, consulting, solution-architecture, or forward-deployed engineering capacity at scale.

Recognised depth in one or more industry domains (BFSI, healthcare/life sciences, retail/CPG, manufacturing, telecom, energy).

Experience running partner-funded or co-innovation programmes with a hyperscaler or model provider (AWS, Microsoft, Google, Anthropic, OpenAI, NVIDIA or ISV partners), including funding programmes, early-access participation, and joint go-to-market.

Certification on a hyperscaler AI/ML or architect track, or equivalent demonstrable platform depth.

Experience with self-hosted / open-weight model deployment (vLLM, Bedrock or Vertex custom endpoints, on-prem GPU) for clients with strict data-boundary requirements.

Published thought leadership, patents, open-source contributions, or conference speaking on GenAI/agentic systems.

has context menu

Mandatory SkillsAI/Gen AI/Gen AI