AI Solutions Architect

About the role

You will work closely with our engineering, product, and architecture teams. Some weeks are whiteboarding sessions and design reviews; others are deep dives into our existing systems.

Duties

• Reviewing our current AI initiatives with the engineering teams -- understanding what is working, identifying consolidation opportunities, and collaborating on a path toward a unified platform
• Working with engineers and product leads to design the reference architecture for multi-agent orchestration, intent classification and routing (including compound/multi-label intents), and how context flows between agents and sessions
• Collaborating on the context management strategy -- token budgets, conversation summarization, scoped context passing between agents, and the tradeoffs between retrieval and compression
• Designing the RAG architecture together with the data and ML teams -- chunking strategies, hybrid retrieval, reranking, citation grounding, and how batch ingestion and real-time serving fit together
• Helping the team establish prompt governance practices -- versioning, A/B testing, performance monitoring, and rollback workflows
• Defining platform resiliency patterns for LLM-dependent systems -- provider failover, circuit breakers, graceful degradation, cost controls, and observability
• Partnering with engineering and product leadership to build a sequenced implementation roadmap that our teams can execute against

Requirements

This is not a wish list. These are the things you will be doing in week one. If you have not done them in production, this is not the right engagement.

• Designed and shipped multi-agent AI platforms -- you know the difference between a demo and a system that handles thousands of concurrent sessions with graceful failure modes
• Built real-time conversational AI systems with proper session memory and context management -- not just chat wrappers around an LLM API
• Architected RAG pipelines that went beyond prototyping -- you have dealt with chunking tradeoffs, embedding drift, stale indexes, and retrieval quality at scale
• Worked across multiple LLM providers (OpenAI, Claude/Bedrock, Gemini, open-source) and understand the real tradeoffs in cost, latency, quality, and reliability -- not just benchmark scores
• Built both real-time and batch ML pipelines and know when to use which -- streaming inference for live interactions, batch processing for catalog-scale operations, and the infrastructure to support both
• Operated in cloud-native environments (AWS, GCP, or Azure) and can make infrastructure decisions, not just architecture diagrams

Preferred Skills/Experience

• Experience in retail, commerce, or customer service AI -- you understand the domain-specific challenges (product catalogs, order state, returns workflows)
• Hands-on with orchestration frameworks (LangGraph, LangChain, LlamaIndex) -- but more importantly, you know their limitations and when to build custom
• Experience with self-hosted model serving (Ollama, vLLM) for cost optimization or data-sensitive workloads
• Have been the person who wrote the AI platform standards that an engineering org of 50+ adopted