AI Solution Architect

Job Description & Summary

At PwC, we help clients build trust and reinvent so they can turn complexity into competitive advantage. We’re a tech-forward, people-empowered network with more than 364,000 people in 136 countries and 137 territories. Across audit and assurance, tax and legal, deals and consulting, we help clients build, accelerate, and sustain momentum. Find out .

Core Responsibilities

Lead technical solutioning in client pre-sales and discovery across all sectors — translating business problems into AI architectures (RAG pipelines, agentic workflows, SDLC automation, data platforms, model risk frameworks)

Own the technical sections of client proposals and engagement scoping documents, including architecture diagrams and implementation sequencing

Build and maintain reusable accelerators and demo assets deployable within 48 hours for client workshops across all use cases

Lead or co-lead technical delivery on AI pilot engagements from architecture through to production handover

Stay current on the AI tooling landscape — with particular depth in the Anthropic/Claude ecosystem — and translate into client-relevant recommendations

Advise on AI governance and responsible AI design, particularly for FS clients subject to MAS regulatory scrutiny on model risk, explainability, and audit trails

Must-Have Skills

Python proficiency — LLM integration, API development, data engineering, and automation scripting

Cloud AI platforms — Azure OpenAI Service, AWS Bedrock, or GCP Vertex AI (at least one in depth)

LLM orchestration — LangChain, LlamaIndex, or equivalent; multi-agent frameworks (CrewAI, AutoGen, or similar)

Vector databases — Pinecone, Weaviate, Chroma, pgvector, or equivalent

Containerisation and CI/CD — Docker, basic Kubernetes, GitHub Actions

5-14 years enterprise technology experience; minimum 2 years in production AI delivery

Anthropic / Claude Ecosystem — Strongly Preferred

Claude API — tool use, computer use, vision, and document processing in production applications

Claude Code — agentic coding workflows, CLI integration, MCP server configuration, and multi-agent software development pipelines

Claude claude.ai and Projects — enterprise deployment patterns, system prompt design, memory and context management

Anthropic prompt engineering — chain-of-thought elicitation, XML-structured outputs, multi-turn conversation design, and retrieval-augmented prompting

Claude model family knowledge — Opus, Sonnet, Haiku trade-offs for latency, cost, and capability in production architectures

MCP (Model Context Protocol) — server implementation, tool registration, and integration with enterprise data sources (Google Drive, Gmail, Slack, CRMs)

Anthropic API batch processing, streaming, and rate limit management for enterprise-scale deployments

AI safety and responsible AI design aligned with Anthropic's principles

Broader AI Tooling — Preferred

Local LLM deployment — Ollama, Qwen, Mistral, Llama on Apple Silicon or equivalent edge hardware for air-gapped or data-sovereign deployments

GitHub Copilot, Cursor, or equivalent AI-assisted development environments — production use in SDLC automation contexts

Open-source agent frameworks — LangGraph, AutoGen, CrewAI, or equivalent for multi-agent orchestration

SDLC and DevTest automation — AI-assisted test generation, code review pipelines, and CI/CD integration

Security and compliance design — data residency, air-gapped deployment patterns, PDPA and MAS regulatory considerations for Singapore deployments

Front-end familiarity — React or equivalent for building lightweight internal tools and executive dashboards

#J-18808-Ljbffr