AVP/AI Engineer

What matters most

• Building Production-Grade Agentic Workflows

  The core of this role is designing and deploying intelligent agents that integrate with enterprise systems. You must demonstrate proficiency in agent orchestration frameworks like LangGraph or AutoGen, focusing on tool integration, memory management, and multi-step workflow execution within a production environment.

• Rigorous AI Observability and Evaluation

  You are expected to implement comprehensive monitoring and evaluation pipelines for AI systems. This includes measuring retrieval relevance, groundedness, and agent performance using frameworks like LangSmith or MLflow to ensure reliability and inform production-readiness decisions.

• Enterprise-Grade Platform Engineering

  The role requires building scalable, secure, and resilient AI infrastructure. You must apply CI/CD, containerization (Kubernetes), and cloud-native security practices to ensure that AI applications are governed, observable, and compliant with enterprise standards.

• AI-Assisted Development and Code Validation

  You will use tools like GitHub Copilot or Cursor to accelerate development. A critical priority is the ability to independently validate AI-generated code for security, functional correctness, performance, and architectural alignment before deployment.

Likely challenges to think through

• Balancing Innovation with Financial Compliance

  Operating in an investment management firm requires strict adherence to security and data governance. The challenge is to innovate with agentic AI while ensuring that data access, model outputs, and agent behaviors meet rigorous financial industry compliance and risk management standards.

• Managing Non-Deterministic Agent Behavior

  Agentic workflows are inherently non-deterministic, which complicates traditional testing. Prepare to discuss strategies for building robust test harnesses, scenario-based simulations, and human-in-the-loop controls to manage failure modes and ensure consistent outcomes in production.

• Integrating AI into Legacy Financial Systems

  You will likely face the difficulty of connecting modern AI agents with established enterprise data platforms and workflows. Prepare to explain how you handle API integration, data quality, and reconciliation when bridging modern AI capabilities with legacy infrastructure.

Topics to prepare

• Agentic Workflow Orchestration

  Focus on the design of multi-step AI workflows, state management, and tool-calling patterns. Exercise: Design a conceptual architecture for an investment research agent that uses multi-step reasoning to retrieve data, perform analysis, and validate outputs against a knowledge graph. This is relevant to the role's requirement for building intelligent agents for portfolio management and risk workflows.

• Production AI Observability and Evaluation

  Study frameworks for monitoring LLM performance, retrieval accuracy, and agent behavior. Exercise: Define a set of metrics to evaluate a RAG system's groundedness and retrieval relevance, and map these to tools like LangSmith or MLflow. This is essential for the role's focus on rigorous validation and production-grade AI system monitoring.

Skills to revise

• Asynchronous Python for AI Systems

  Develop proficiency in writing non-blocking, asynchronous Python code for high-throughput AI applications. Exercise: Refactor a synchronous API call sequence into an asynchronous pattern using 'asyncio' to handle concurrent model inference requests. This is critical for the role's requirement to build scalable, production-grade AI applications.

• Containerized Deployment and CI/CD

  Practice deploying AI models and services using Docker and Kubernetes. Exercise: Create a Dockerfile for a model inference service (e.g., using vLLM) and define a basic CI/CD pipeline configuration that includes automated testing and security scanning. This aligns with the role's emphasis on platform engineering and governed deployment.

• RAG Architecture Implementation

  Gain hands-on experience with hybrid retrieval techniques and metadata filtering. Exercise: Build a small-scale RAG pipeline that implements both vector search and metadata-filtered retrieval to improve query precision. This is directly requested for the firm's next-generation enterprise AI platform.

• Architect an Agentic RAG Workflow

  Prepare a technical walkthrough of a RAG system you have built, specifically highlighting how you implemented agent orchestration, tool calling, and memory management. Be ready to explain your choice of framework (e.g., LangGraph or AutoGen) and how you handled multi-step reasoning.

• Demonstrate Production-Grade Observability

  Prepare to discuss how you have implemented monitoring for AI systems. Focus on your experience with frameworks like LangSmith or MLflow to track retrieval relevance, model latency, and agent success rates in a production environment.

• Showcase Platform Engineering Proficiency

  Create a summary of your experience deploying AI models within containerized environments (Docker/Kubernetes). Emphasize your approach to CI/CD pipelines, secrets management, and ensuring security in the software supply chain.

• Prepare for AI-Assisted Development Validation

  Develop a strategy for independently validating code generated by AI tools like GitHub Copilot or Cursor. Be prepared to explain your methodology for checking functional correctness, security vulnerabilities, and performance bottlenecks.

• Formulate Strategic Questions for the Team

  Prepare thoughtful questions regarding the firm's approach to model governance, how they handle human-in-the-loop controls for high-stakes investment workflows, and how the team balances commercial model usage with self-hosted open-source alternatives.