From Prepayments to Pricing: VP AI Data Scientist Driving MBS/ABS Models in NYC

AI/ML Data Scientist, Fixed-Income Structured Products - VP (New York)

This role sits at the intersection of deep learning and structured finance. If you build or ship products in fixed-income, this is the kind of hire that moves the needle: models that improve prepayment and default forecasts, valuation accuracy, and the speed of decision-making in MBS and ABS markets.

The team is high-performing, the mandate is clear, and the expectations are senior. You'll lead research, build production systems, and turn raw data into front-line signals used by portfolio managers and risk teams.

Role Snapshot

• Title: AI/ML Data Scientist, Fixed-Income Structured Products - VP
• Company: Selby Jennings (on behalf of a high-performing Mortgage Quant team)
• Location: New York (Manhattan, United States)
• Type: Full-time / Hybrid, Permanent
• Compensation: USD 200,000-250,000 base (approx. USD 325,000-375,000 total comp)

What You'll Lead

• AI and ML research across mortgage and structured product strategies.
• Predictive models for prepayment, default, and valuation across MBS and ABS.
• Deep learning and transformer architectures to lift forecast accuracy and risk assessment.
• Scalable multi-agent or automated systems for valuation workflows and investment analytics.
• NLP-driven tools: AI news analytics, LLM-based apps, and market signal generation platforms.
• Hands-on collaboration with PMs, quants, and engineering to ship production systems.
• Active tracking of advances in deep learning and language models relevant to markets.

Qualifications That Fit

• Ph.D. or Master's in Computer/Electrical Engineering, Computer Science, Applied Math, Physics, or similar. Focus on deep learning preferred.
• Undergraduate foundation in applied math, physics, engineering, or comparable quantitative paths.
• Experience in fixed-income analytics with exposure to mortgages, structured products, or related asset classes.
• Proven record with transformer-based, deep learning, or hybrid ML systems for forecasting, valuation, or risk modeling.
• Track record building NLP, LLM, or QA systems for real financial or enterprise use cases.
• Production deployment experience with cross-functional stakeholders.
• Background with financial institutions, hedge funds, or ML centers of excellence is a plus.

Technical Stack

• Programming: Python, SQL
• Frameworks & Tools: PyTorch, TensorFlow, Hugging Face, LangChain, and related ML/NLP libraries
• Core Skills: Model training and optimization, data engineering workflows, cloud or distributed computing

Why This Role Matters (Finance & Product Perspective)

• Alpha and carry with guardrails: Better prepay and default modeling tightens pricing, hedging, and scenario analysis-small error reductions compound into real P&L.
• Faster delivery to desks: Automated valuation and LLM-driven signal pipelines shorten time from idea to trade. Lower latency, fewer manual steps.
• Repeatable decisions: Standardized ML workflows with versioned data and models reduce variance across desks and support model governance.
• Explainability where it counts: Methods for feature attribution, challenger models, and policy checks help pass review with risk and compliance.
• Product thinking: This isn't just research. It's scoped like a product-user needs (PMs/analysts), SLAs, dashboards, and measured outcomes.

How to Scope the Work (Practical Guide)

• Use cases: Prepayment and default forecasting, loan-level stratification, pool selection, BWIC support, scenario valuation, and macro-news to spread mapping via NLP.
• KPIs: Out-of-sample WAPE/MAE for pool-level prepay, default AUC/KS, fair value hit rate vs. market prints, latency per valuation, and model drift frequency.
• Risk controls: Backtests with regime buckets, challenger/baseline comparisons, and automated monitoring for data shifts and performance decay.
• Delivery: APIs for desk tools, notebooks for research, and dashboards that PMs can trust at 8:30 a.m.

What "Great" Looks Like in 90 Days

• Days 0-30: Reproduce baseline models, map data lineage (loan-level, pool-level, macro, servicing), and align with PMs on what "better" means in dollar terms.
• Days 31-60: Ship a transformer-based prototype on a priority use case (e.g., prepay on specified pools). Stand up a simple valuation API and monitoring.
• Days 61-90: Integrate with production workflows, roll out dashboards, run side-by-side P&L attribution vs. prior models, and set thresholds for alerts and retrains.

Role Details (At a Glance)

• Team: Mortgage Quant team focused on structured products and fixed income
• Focus: End-to-end research, model development, and production deployment
• Job ID: PR/572260

About the Company

Selby Jennings - New York, United States - ~1000 employees - HR & Recruitment. They support the Financial Sciences & Services industry with talent that can shape the future of a business.

Helpful Context

• Federal Reserve: Mortgage-Backed Securities (overview)
• BIS: AI and Machine Learning in Finance (summary)

Level Up Your Team's Skills

• AI tools for finance: curated picks for research, risk, and reporting
• Courses by job: targeted learning paths for data scientists and PMs

Final Notes

This role is built for someone who can own the full stack: research, data, models, deployment, and stakeholder buy-in. If you're a finance or product leader, expect impact measured in tighter pricing, faster workflows, and clearer signals to the desk. If you're the candidate, bring depth in deep learning and the pragmatism to ship.