Job Description & Details
This is a heavy-hitting architecture role focused entirely on solving the state and memory problem for enterprise AI agents. If you are tired of building basic RAG wrappers and want to design foundational context-passing, memory architectures, and agentic workflows at scale, this is where you do it. You will be defining how large enterprise domains like wealth management and customer service maintain continuous, trustworthy context across complex interactions.
What You'll Actually Be Doing
Your days will be split between designing enterprise-grade context pipelines and sparring with product and security teams. You'll be building frameworks that decide what an LLM actually remembers between turns, optimizing context windows so you aren't burning tokens on garbage data, and stitching together vector databases with structured enterprise systems. You'll also spend significant time setting up multi-agent state-sharing, ensuring that when one AI handoff happens to another, the context doesn't get lost in translation.
The Core Tech Stack
Azure is your home base here, so deep cloud-native chops are non-negotiable. You need to live and breathe Python, Java, or C#, paired with modern AI orchestration tools like LangChain, LangGraph, Semantic Kernel, or LlamaIndex. Vector databases (Pinecone, Weaviate, Azure AI Search) and graph databases are critical for the retrieval and knowledge-grounding pieces. You won't just be calling APIs; you'll be building the distributed microservices and event-driven pipelines that feed these models reliably.
Interview Expectations
Expect the hiring panel to test your design chops with an open-ended system prompt challenge, like asking you to architect a low-latency episodic and semantic memory framework for a multi-agent financial advisory workflow. They want to see how you handle context compression, token optimization, and hallucination reduction at scale. Another favorite line of questioning will test your approach to enterprise security, specifically how you plan to scrub PII and sensitive financial data out of long-term vector stores without breaking semantic retrieval.
Application Advice
Skip the generic resume summary and put your distributed systems and Generative AI platform wins right at the top. Make sure exact keywords like Azure, RAG, vector databases, LangGraph, and event-driven architectures are clearly visible in your experience bullets to clear the ATS. If you have built custom context caching, semantic retrieval pipelines, or multi-agent systems in production, highlight those metrics front and center.