Enterprise AI Architect

Enterprise AI Architect JD

What You Will Do

• Design and implement context architectures for LLM apps and agents: schemas, memory patterns, context assembly, and context window management.
• Build and optimize RAG pipelines (chunking, embeddings, hybrid retrieval, re-ranking) and validate quality using repeatable evaluation harnesses.
• Develop system prompts, prompt libraries, and structured output patterns; harden solutions against prompt injection and jailbreak attempts.
• Implement agentic workflows and tool-use integrations (APIs, function calling, workflow engines) with clear guardrails and observability.
• Engineer memory and persistence patterns (session memory, episodic recall, vector memory) appropriate for enterprise privacy and retention needs.
• Work with enterprise data and app teams to connect AI solutions to real systems (ERP/SCM/CRM, data lakes/warehouses), ensuring secure access and correct semantics.
• Collaborate with delivery leads to break work into stories, estimate effort, and drive day-to-day execution; mentor engineers through reviews and pairing.
• Use systematic optimization (prompt/context tuning, retrieval experiments, DSPy-style approaches) to improve reliability, latency, and cost.
  
  

• Strong Python skills and ability to ship production services.
• Hands-on expertise with RAG: embeddings, vector stores, retrieval strategies, re-ranking, and grounding techniques.
• Strong prompt and context engineering: system prompts, structured outputs, tool-use prompting, and context assembly patterns.
• Experience building agentic systems with orchestration frameworks (or custom implementations) and designing safe tool integrations.
• Awareness of security threats (prompt injection, data exfiltration) and ability to implement practical mitigations and guardrails.
  
  

• Experience integrating AI apps with enterprise services, data sources, and identity (SSO/IAM), including secure network and secrets handling.
• Ability to work with structured and unstructured enterprise data; understand governance/lineage enough to avoid incorrect or unsafe data use.
• Comfort operating within enterprise SDLC controls: CI/CD, change management, security reviews, and production incident response.
• Working knowledge of enterprise workflows and process context so AI solutions map to real operations and decision points.
  
  

• Python; common LLM/RAG frameworks (LangChain, LlamaIndex, Haystack or equivalent).
• Vector databases and search stacks (pgvector, Pinecone, Weaviate, Milvus, Elasticsearch/OpenSearch).
• Memory and state management approaches (session stores, vector memory, durable stores) appropriate for privacy and retention constraints.
• Evaluation and observability tooling (RAGAS, LangSmith/Phoenix or equivalent) and ability to build custom eval pipelines.
  
  

• B.Tech / M.Tech in CS, Engineering, or Linguistics; research background in NLP or information retrieval is a plus.
• Published work, open-source contributions, or internal frameworks related to context management or prompt engineering.
• Prior consulting or professional services experience — ability to adapt context design to diverse client environments quickly