AI Architect

HRG Group; Pune District; Vor Ort; INR 4.000.000 - 10.000.000; Vollzeit

Zusammenfassung

Darwinbox Digital Solutions Pvt. Ltd. in Pune invites an AI Architect to lead design of AI-assisted and AI-native solutions for digital platforms. The role blends engineering, product, and client-facing work, with hands-on development and governance responsibilities.

You will translate complex problems into architecture, guide patterns such as RAG and MCP, and drive robust estimations, delivery approaches, and security considerations across the stack.

Qualifikationen

• 12+ years in software engineering and architecture.
• 4+ years delivering LLM/GenAI in production.

Aufgaben

• Translate business problems into AI-ready architectures for publishing/learning products.
• Choose model patterns (prompt/workflow, RAG, fine-tuning, agents/MCP) and justify build-vs-buy decisions.
• Provide rapid, defendable estimates and delivery approaches for proposals.
• Support pre-sales with architecture walkthroughs for engineers to C-suite.
• Develop proofs of concept and reference implementations to de-risk delivery.
• Define evaluation harnesses, guardrails, and acceptance criteria for AI features.
• Create reusable accelerators and an AI engineering playbook across teams.
• Mentor engineers on AI development with emphasis on quality, security, and cost.

Kenntnisse

AI architecture

LLM/GenAI

Cloud platforms

RAG

MLOps

Security

Prototyping

Pre-sales

Tools

AWS SageMaker

Azure ML

Google Vertex AI

MLflow

Kubeflow

Weights & Biases

Jobbeschreibung

Designation: AI Architect — Digital Publishing

Office Location:Pune

Role overview

We're looking for an AI Architect having more than 12 years of experience to lead the design of AI-assisted and AI-native solutions across our digital platforms. Architect will turn open-ended problems into clear architectures, working prototypes, and credible estimates — quickly — while owning the risk, security, and governance that come with putting AI into content- and IP-heavy products. It's a hands-on, high-influence role that sits between engineering, product, and client-facing teams.

Responsibilities

Solutioning & architecture

Translate ambiguous business problems and RFPs into target architectures, trade-off analyses, and reference designs for AI-enabled publishing and learning products.

Pick the right pattern for each problem — prompt/workflow, RAG, fine-tuning, agents/MCP, or classic ML — and justify build-vs-buy and model choices.

Produce fast, defensible effort estimates, delivery approaches, and team shapes for proposals and new initiatives.

Support pre-sales and client conversations with architecture walkthroughs pitched anywhere from engineer to C-suite.

Hands-on AI build & POCs

Stand up proofs of concept and reference implementations that de-risk the hard parts before full delivery commits.

Define evaluation harnesses, guardrails, and acceptance criteria so AI features ship with measurable quality.

Establish reusable accelerators, patterns, and an AI engineering playbook across teams.

Mentor engineers on AI-assisted development and raise the bar on quality, security, and cost awareness.

We mean two different things here, and we expect both.

AI-assisted development — you use AI across the SDLC and lift team velocity with it. You're fluent with AI coding assistants and agentic developer workflows, you prompt and review their output critically, and you know where they genuinely help versus where they create rework or risk.

AI-native development — you design systems where LLMs, retrieval, and agents are first-class components, not bolt-ons. That spans orchestration and tool use (including MCP), retrieval and grounding, evaluation and observability, guardrails, prompt and context management, and the cost and latency engineering these systems need to survive production.

You’ll architect and build across our core stack, so hands-on fluency in most of these is expected:

Cloud: Azure-based cloud development and services, Cloud Native development

Deep experience with cloud ecosystems like Amazon Web Services (AWS) (Amazon SageMaker), Microsoft Azure (Azure ML), or Google Cloud Platform (GCP) (Google Vertex AI).

Familiarity with MLOps and orchestration tools such as MLflow, Kubeflow, or Weights & Biases.

Responsible AI

You treat AI as a capability to be governed, not just adopted.

Risk & mitigation — you can name and design around the failure modes of AI usage: hallucination and grounding failure, IP and content leakage, data governance and privacy, evaluation gaps, runaway cost and latency, and model or vendor lock-in. You build in human-in-the-loop checkpoints, evals, and fallbacks rather than trusting the model to behave.

Security — you understand the security surface that AI and agents introduce, and how it differs from traditional app security: prompt injection (direct and indirect), retrieval that authorizes before it retrieves rather than after, agent and tool permissioning, MCP scopes and confused-deputy risks, tenant isolation, and audit trails of what the model saw, decided, and did.

In a publishing context you also think about content and IP exposure — what a legitimate account, automated user, or agent can extract from the platform, and how that extraction would be detected and traced.

What you bring

12+ years in software engineering and architecture, including 4+ years designing or delivering LLM/GenAI solutions in production (tune the years to your seniority bar).

Strong grounding in cloud (AWS, Azure, or GCP), modern application and data architecture, and API/integration design.

Practical depth in several of: RAG and vector search, orchestration/agent frameworks, prompt and context engineering, evaluation tooling, and MLOps/LLMOps.

Security fundamentals for web and AI systems (authentication/authorization, data protection, secure SDLC) plus a working command of AI-specific threats.

A track record of estimating, scoping, and sequencing delivery, and of communicating clearly with both engineers and executives.

Comfortable being hands-on in code and prototypes, not only on slides.

Digital publishing or eLearning domain familiarity is a strong plus: content platforms and headless CMS, SCORM/xAPI, DRM and content protection, metadata, rights, and accessibility.

Experience with MCP, multi-agent systems, or agent evaluation and safety tooling.

Publishing and learning standards: EPUB, ONIX, SCORM/xAPI, QTI, IMS/LTI.

Relevant cloud or AI certifications, and recognized security credentials.

Open-source contributions, writing, or speaking on applied AI or AI security.