AI Engineer

Demonstrated ability to design state graphs, conditional edges, and multi-agent compositions.\nModel Context Protocol experience required. Comfortable designing tool calls and building protocol wrappers. Agentic pair programming with generative AI as the primary working mode.

Prior experience is required.\nPydantic for data validation and structured outputs across agent systems and APIs. SQL proficiency and PostgreSQL experience.\nVector database experience (pgvector, Azure AI Search, or similar). Familiarity with modern data platforms such as Snowflake and Databricks.\nMulti-step agent systems with proper evaluation and validation.\nStrong fluency across modern frontier language models such as the GPT, Claude, and Gemini model families.

Model selection is driven by the requirements of each task and the practical considerations of cost.\nThe discipline to choose mature, proven frameworks for AI engineering work. Technical decisions are grounded in clear engineering justification and supported by data. Docker and containerization for development and deployment workflows.\nFastAPI or equivalent web frameworks for building APIs and backend services.

Working knowledge of GitHub workflows, code review discipline, and infrastructure-as-code patterns\nAlthough Microsoft Azure is the primary platform, Google Cloud Platform experience is also valuable and applicable to the practice's work.\nFamiliarity with Azure Container Apps, Azure Kubernetes Services, or similar container orchestration platforms.\nFamiliarity with Apache Software Foundation tools such as Apache Airflow, Apache Kafka, and Apache Flink. Experience with LangChain, PyTorch, or other AI and machine learning frameworks.\nCrewAI or other multi-agent frameworks. LlamaIndex for RAG and data ingestion workflows.\nCelery and Redis for background task processing and caching.

Familiarity with A2A (Agent-to-Agent) protocol for inter-agent communication.\nExperience with modern natural language processing tools, including embedding models and entity recognition.\nFamiliarity with vision-language model integration for multi-modal AI use cases. Experience working in regulated markets, including the compliance and risk-management disciplines such environments require.\nPlaywright for end-to-end testing automation.\nFamiliarity with hybrid architectures that integrate deterministic and generative AI techniques.