#CustomAI

Custom AI

**Customer S&S Senior Consultant**

About the Role

Deloitte is seeking a Full Stack CRM Architect to own the design and delivery of components of custom, AI-augmented Customer Relationship Management platforms. This role sits between hands-on engineering and solution architecture: you will lead module- and feature-level design decisions, build critical parts of the platform yourself, and provide day-to-day technical direction to a small delivery pod.

You will partner closely with Managers and Senior Managers/Architects to translate the engagement’s overall architectural vision into working software, while mentoring Consultants and helping instill Deloitte’s engineering standards on the ground. This is a strong next step for engineers ready to move from building features to owning the design of the systems those features live in.

Key Responsibilities

1. Architecture & Solution Design (Module Ownership)

• Own the detailed design of specific modules or services within custom CRM platforms — including data models, API contracts, and front-end structures — in line with the overall architecture set by the Manager/Architect. May approve module-level design changes independently; changes affecting the overall architecture pattern require Architect/Manager sign-off.

• Architect and build full-stack features using modern frameworks (React, Angular, Next.js) for frontend and Node.js or Python/FastAPI for backend, with an eye toward scalability and maintainability.

• Apply and reinforce architecture governance standards and reusable microservice patterns within your workstream; identify where existing patterns should be extended.

• Contribute technical analysis to build vs. buy decisions, including prototyping and comparative evaluation of candidate approaches.

2. Product & Delivery Leadership

• Co-lead discovery sprints and user feedback sessions for your workstream; translate business pain points into well-formed backlog items and acceptance criteria.

• Track adoption and business-outcome metrics for the features your pod delivers, not just delivery dates.

• Help run innovation labs and internal hackathons, guiding Consultants through prototyping emerging CRM capabilities (AI agents, voice interfaces, personalization engines).

• Build and help publish reusable components toward Deloitte’s proprietary CRM accelerators and reference architectures.

• Research emerging technologies (agentic frameworks, LLM orchestration tools, edge inference) and recommend where they apply to current engagements.

3. AI, GenAI & Agentic Integration

• Design and build AI/GenAI features within CRM workflows — including lead-scoring models, predictive components, agentic workflow steps, and intelligent content generation.

• Build RAG pipelines and LLM orchestration components (LangChain, LlamaIndex) grounded in CRM knowledge bases; implement semantic caching and query routing to reduce inference costs.

• Apply AI cost-governance practices within your workstream: model cascading (routing simple queries to smaller models), token budget tracking, and prompt compression.

• Deploy and tune inference-serving components (containerized vLLM, managed inference endpoints), applying batching and caching strategies under senior guidance.

• Build conversational AI features (chatbots, virtual agents) that automate Tier-1 service cases, meeting latency SLAs (<500ms P95) via request batching and streaming.

• Implement output-validation and hallucination-detection checks for GenAI features you own, escalating drift or quality concerns to the Architect/Manager.

• Track AI-attributed metrics (pipeline velocity contribution, cost-per-interaction, model efficiency) for your workstream and feed them into program-level reporting.

4. AI Implementation Imperatives — Applied at the Workstream Level

Area

What this level does

Data First

Implement data-quality pipelines and CRM hygiene checks (deduplication, completeness scoring, freshness SLAs) for your module ahead of any AI deployment.

Scalability by Design

Design your module’s AI-serving components to scale horizontally from day one (autoscaling, async queues) in line with the program’s overall pattern.

Cost Efficiency

Right-size models to task complexity within your workstream; implement semantic caching and monitor token spend against budget for the features you own.

**Adaptable Archit…