What you will do:

• Design and implement AI agents capable of planning, reasoning, and executing multi-step workflows
• Build single-agent and multi-agent architectures for enterprise use cases
• Develop reusable agent patterns and templates that other teams can build on
• Deploy agents on managed platforms (Anthropic, Azure OpenAI, GCP Vertex AI)
• Configure secure execution environments, lifecycle management, and session handling
• Design Human-in-the-loop escalations and approval gates for high value actions
• Own the agent release lifecycle: prompt and config versioning, staged rollouts, and canary releases
• Integrate agents with our internal data platforms, APIs, and services
• Implement RAG pipelines, vector databases, and structured knowledge layers
• Build evaluation frameworks measuring accuracy, task completion, and reliability
• Monitor agent behavior and performance in production
• Implement fallback mechanisms, error handling, cost controls, and usage guardrails
• Ensure agents operate within our security policies, data access constraints, and regulatory frameworks
• Maintain audit trails and explainability of agent actions
• Deliver solutions covering investigation workflows, reporting, reconciliation, and decision support
• Implement safeguards against hallucinations, unsafe actions, and data leakage
• Translate workflows from Trading, Risk, Compliance, Finance, and Operations into agent-driven automation

What you need to succeed: 

• Strong programming skills in Python
• Hands-on experience with LLM-based applications and agent systems
• Familiarity with agent frameworks such as LangChain, LangGraph, AutoGen, or CrewAI
• Experience with API development, microservices, and tool orchestration
• Knowledge of vector databases, RAG pipelines, and prompt engineering
• Experience with cloud platforms (AWS, Azure, or GCP)
• Containerization (Docker) and basic orchestration (Kubernetes)
• CI/CD pipelines, production deployment, and observability tooling
• Ability to design goal-driven AI systems rather than static models
• Experience building multi-step reasoning workflows
• Ability to implement evaluation pipelines and production guardrails
• Awareness of cost, latency, and performance trade-offs in LLM systems