BNY - AI Data Science Specialist - Markets Agentic AI & Model Governance

Role Summary:

This role enables productiongrade, governed AI capabilities for Markets by designing, validating, and operating AI agents across their full lifecycle. The individual owns prompt engineering, model selection, RAG/KAG architectures, and ongoing model governance, ensuring AI solutions are explainable, stable, and compliant with DURB and MRMG expectations.

This is a continuous BUILD + RUN role, not a onetime implementation.

Key Responsibilities:

BUILD Design & Delivery:

• Define agent logic, behavior, and boundaries for Markets use cases (risk, trading, credit, surveillance, analytics).

• Design stable, testable, and explainable prompts, with versioning and regression testing.

• Select, justify, and document AI / ML models appropriate to usecase criticality and risk tier.

Design AI architectures, including:

• RetrievalAugmented Generation (RAG)

• KnowledgeAugmented Generation (KAG)

- AgenttoAgent communication patterns

• MCPstyle server architectures for tool orchestration and control

• Establish validation and testing strategies so outputs are consistent, trusted, and auditable.

• Produce MRMG / DURBready artefacts (methodology, data lineage, controls, fallback plans) in parallel with development.

RUN Operate & Evolve:

• Monitor model and agent behavior in production, including drift, instability, and control breaches.

• Continuously tune prompts, retrieval logic, and models as:

• New datasets are onboarded

• Market conditions shift

- Risk and regulatory expectations evolve

• Own ongoing performance monitoring (OPM) and respond to governance reviews, findings, and control enhancements.

• Ensure AI agents remain assistive, bounded, and compliant over time.

Required Technical Skills:

AI / ML Models:

• Classical & Statistical ML:

• XGBoost / Gradient Boosting

• Timeseries models

Unsupervised Learning:

- Clustering (kmeans, hierarchical, DBSCAN)

• Deep Learning & NLP:

• Transformer based models

• Large Language Models (LLMs)

• Embedding models for semantic search

• Strong understanding of model strengths, limitations, and failure modes.

AI Architecture & Engineering

• Retrieval systems: RAG, KAG, semantic search, vector databases.

• Agentic systems:

• Agent orchestration

• Agent to agent communication

• Tool calling and decision gating

• MCP style server architectures for secure tool access, observability, and control.

• Model versioning, prompt versioning, and experiment tracking.

Governance & Risk:

• Hands on experience supporting DURB and MRMG approvals.

• Ability to translate technical designs into governance ready documentation.

• Strong understanding of:

• Model risk management

• Explainability and auditability

• MNPI, information barriers, and fiduciary risk

• Comfortable operating under regulatory scrutiny and evolving control frameworks.

Markets & Business Context:

• Strong Markets domain awareness (products, workflows, risk sensitivities).

• Ability to partner with Front Office, Risk, Technology, and Governance stakeholders.

• Capable of balancing delivery speed with control rigor.

Why This Role Matters

AI in Markets does not self govern. Prompt quality, model choice, architecture, and validation directly impact market integrity, risk outcomes, and regulatory posture. This role ensures AI remains trusted, controlled, and scalable as adoption grows.