Description

- Building and extending LLM-powered agents in Python: intent classification, domain-specialist agents, and agentic extraction flows that decompose multi-part user questions

- Natural-language-to-structured-payload pipelines: parsing user utterances into JSON query payloads (filters, exclusions, rankings, metric selection) with high precision

- Integrating semantic search / vector retrieval services (embedding-based entity resolution, fuzzy matching, confidence thresholds, disambiguation flows)

- Building and consuming FastAPI microservices; async orchestration of parallel LLM and API calls

- Prompt engineering, structured output enforcement (JSON schema / function calling / tool use), and guardrails

- Evaluation harnesses: building test sets, measuring extraction accuracy, regression-testing prompt and model changes

- Working with metadata/catalog services, entitlement-aware data access, and reporting-engine payload contracts

- Collaborating across multiple service teams; writing clear technical documentation (Confluence, ADRs, sequence/flow diagrams)

MUST-HAVE SKILLS

- 5+ years of professional Python — advanced level; clean, tested, production-grade code (typing, pytest, packaging, code review discipline)

- 1.5+ years hands-on building GenAI/LLM applications in production (not POCs only): OpenAI / Anthropic / Azure OpenAI / Bedrock or similar APIs

- Agentic frameworks and patterns: LangChain/LangGraph, LlamaIndex, or equivalent hand-rolled orchestration; tool/function calling; multi-step agent flows

- Structured output extraction from LLMs: JSON schema enforcement, Pydantic, retry/repair strategies

- RAG and vector search: embeddings, chunking strategies, hybrid search, reranking (any of pgvector, Pinecone, Weaviate, OpenSearch, or warehouse-native vector functions)

- FastAPI (or Flask/Django with strong API design), async Python, REST integration patterns

- SQL proficiency and comfort working against large analytical datasets

- Git, CI/CD, Docker; comfortable in cloud environments (Azure preferred; AWS/GCP acceptable)

- Strong communication

 — these roles interact directly with multiple engineering teams and product owners NICE-TO-HAVE

- Snowflake (especially Cortex functions / Snowpark) - Experience with NL2SQL or NL-to-query-DSL systems

- Retail / CPG / market-measurement data domain exposure

- MCP (Model Context Protocol) or similar tool-integration standards

- LLM evaluation tooling (Ragas, promptfoo, custom eval harnesses); observability (LangSmith, Langfuse)

- Prior experience in multi-agent systems with disambiguation/human-in-the-loop flows

Key Skills

Education

Bachelor's degree