As a Generative AI Agent Developer, your primary responsibilities will include:
Agent Orchestration: Design, develop, and deploy complex, multi-step AI agents using LangChain, LangGraph, and ADK, ensuring efficient state management and execution paths.
Tool and Function Calling: Implement robust mechanisms for agents to utilize external APIs, custom Python functions, and specialized tools (Tool Use) to achieve complex goals and interact with enterprise systems.
GCP Service Integration: Leverage the full suite of Vertex AI services (Agent Builder, Pipelines, Workbench) to manage the entire agent lifecycle, from testing to scalable production deployment.
Model Integration and Customization: Integrate and manage various foundational models (GCP, open source) and utilize Vertex AI Training for model customization where required.
Infrastructure & Networking: Utilize general GCP infrastructure knowledge, including Compute (Agent Engine, GKE, Cloud Run) for deployment, and adhere to networking standards (VPC, Load Balancing, Cloud DNS).
Performance and Reliability: Implement rigorous testing and logging to monitor agent behavior, optimize token usage, ensure reasoning accuracy, and minimize potential hallucinations.
Documentation & Best Practices: Document agent architectures, development processes, and promote reusable design patterns across the engineering team.
Nice-to-Have Skills
While not strictly required, candidates possessing the following skills will have a distinct advantage:
Graph Databases: Experience with graph databases (e.g., Neo4j) and their integration into LangGraph or RAG systems to manage complex relationships and contextual memory.
Low-Latency Deployment: Experience optimizing GenAI applications for low-latency, high-throughput environments (e.g., using quantization or compiling models).
Security & Compliance: Familiarity with data security principles related to PII/PHI handling within GenAI pipelines and implementing filtering mechanisms (e.g., DLP).
Advanced MLOps: Deep experience with running production workloads on GKE or leveraging advanced features of Cloud Run for model serving.
Agent Orchestration: Design, develop, and deploy complex, multi-step AI agents using LangChain, LangGraph, and ADK, ensuring efficient state management and execution paths.
Tool and Function Calling: Implement robust mechanisms for agents to utilize external APIs, custom Python functions, and specialized tools (Tool Use) to achieve complex goals and interact with enterprise systems.
GCP Service Integration: Leverage the full suite of Vertex AI services (Agent Builder, Pipelines, Workbench) to manage the entire agent lifecycle, from testing to scalable production deployment.
Model Integration and Customization: Integrate and manage various foundational models (GCP, open source) and utilize Vertex AI Training for model customization where required.
Infrastructure & Networking: Utilize general GCP infrastructure knowledge, including Compute (Agent Engine, GKE, Cloud Run) for deployment, and adhere to networking standards (VPC, Load Balancing, Cloud DNS).
Performance and Reliability: Implement rigorous testing and logging to monitor agent behavior, optimize token usage, ensure reasoning accuracy, and minimize potential hallucinations.
Documentation & Best Practices: Document agent architectures, development processes, and promote reusable design patterns across the engineering team.
Nice-to-Have Skills
While not strictly required, candidates possessing the following skills will have a distinct advantage:
Graph Databases: Experience with graph databases (e.g., Neo4j) and their integration into LangGraph or RAG systems to manage complex relationships and contextual memory.
Low-Latency Deployment: Experience optimizing GenAI applications for low-latency, high-throughput environments (e.g., using quantization or compiling models).
Security & Compliance: Familiarity with data security principles related to PII/PHI handling within GenAI pipelines and implementing filtering mechanisms (e.g., DLP).
Advanced MLOps: Deep experience with running production workloads on GKE or leveraging advanced features of Cloud Run for model serving.
