AI Engineer- Traded Risk

We are currently seeking an experienced professional to join our team in the role of AI Engineer- Traded Risk (FTC role).

In this role, you will:

• Design and implement single-agent and simple multi-agent workflows using frameworks such as LangGraph, LangChain, AutoGen, CrewAI, etc.
• Lead the end-to-end development of Agentic AI applications, from ideation and data exploration to rapid prototyping and initial deployment.
• Build and maintain tool integrations (function calling / MCP servers) connecting LLM agents to internal risk systems.
• Implement RAG (retrieval-augmented generation) pipelines over risk policy documents, regulatory guidance, and historical incident logs to ground agent outputs in verified sources.
• Write and iterate on prompt templates, system instructions, and few-shot examples, and maintain a versioned prompt library.
• Build evaluation harnesses to test agent accuracy, hallucination rate, and task completion, using both automated metrics and structured human review.
• Implement guardrails: input/output validation, PII and confidential-data filtering, approval checkpoints for any agent action with financial or regulatory consequence.
• Support human-in-the-loop design , ensuring every agentic workflow has a clear escalation path when confidence is low or an anomaly is detected.
• Monitor deployed agents in production: track cost (token usage), latency, failure modes, and drift in output quality over time.
• Document workflows, decision logic, and control points clearly enough for audit and model-risk review.
• Collaborate with market risk SMEs to translate manual, judgment-heavy processes into structured agent tasks.

To be successful you will:

• Have working proficiency in Python, including API integration, async programming, and data manipulation (pandas/numpy).
• Have hands-on experience building agentic applications , this can come from personal projects, hackathons, internships, or professional work.
• Have practical exposure to at least one agent orchestration framework (LangGraph, LangChain, CrewAI, AutoGen, Semantic Kernel) or direct experience with the Claude/OpenAI/Gemini APIs including function calling and tool use.
• Understand prompt engineering fundamentals: system prompts, chain-of-thought/structured reasoning prompts, few-shot examples, output formatting (JSON schemas, XML tags).
• Be familiar with RAG architectures: embeddings, vector databases (e.g., Pinecone, Weaviate, FAISS, pgvector), chunking strategies, and retrieval evaluation.
• Have basic understanding of LLM evaluation techniques , golden datasets, LLM-as-judge patterns, regression testing for prompt changes.
• Have foundational understanding of market risk concepts , VaR, Greeks, stress testing, limit frameworks , gained through coursework, self-study, or prior exposure.
• Hold FRM Part I/II, CQF, or equivalent certification in progress or completed.
• Have experience with MCP (Model Context Protocol) server development for tool integration.
• Have exposure to model risk management (MRM) or SR 11-7-style validation frameworks.
• Have prior experience in a regulated environment (banking, insurance, asset management).
• Have a background in software engineering, data engineering, or quant development, given the technical build nature of this role.