Senior ML Engineer

Title:
 Senior Machine Learning Engineer / Applied Scientist – Mortgage AGI

Location:
 Remote (US)

Company:
 LoanPASS

About LoanPASS

LoanPASS is a modern, fully configurable, no-code rules engine and pricing platform for lenders. Over the past 8 years, LoanPASS has been built to do something no other mortgage technology platform has done: 
atomically unify underwriting, pricing, and locking
 in one deterministic engine.

LoanPASS is now extending that foundation into a new frontier: a 
Mortgage AGI
 layer that learns from millions of AUS decisions and historical loan performance to help lenders design, price, and manage their credit boxes in real time. This role is the founding ML hire for that initiative.

The Role

LoanPASS is looking for a 
Senior Machine Learning Engineer / Applied Scientist
 to lead the modeling work behind its Mortgage AGI initiative.

You will take labeled decisions from the LoanPASS AUS (millions of synthetic scenarios plus historical performance) and turn them into:

• Credit risk models (PD / LGD / prepay),
• AUS behavior approximators where needed,
• Portfolio and credit-box optimization logic (volume vs. ROE vs. risk).

You'll work closely with engineering, product, and credit leaders to turn the LoanPASS rules engine into a 
learning system
 that continuously improves lender credit policy, not just enforces it.

What You'll Do

Modeling & Research

• Design and build models for:
• Default probability (PD),
• Loss given default (LGD) / severity,
• Prepayment / CPR,
• Potentially AUS emulators for low-latency cases.
• Combine 
  AUS outputs
  , 
  FICO 10T
  , 
  loan attributes
  , 
  HPA
  , and 
  macro variables
   into robust training datasets.
• Use the right tools for the job:
• Tree-based models (XGBoost / LightGBM / CatBoost),
• Survival/competing risk models for default/prepay,
• Simple time-series models where appropriate.

Data & Pipelines

• Define the data contracts for:
• Synthetic scenarios generated by the AUS,
• Historical agency / lender loan tapes,
• Macroeconomic overlays (rates, unemployment, HPA by market, etc.).
• Build and own the training pipelines:
• Feature engineering,
• Train/validation/test splits and regime-aware evaluation,
• Versioning, experiment tracking, and documentation.
• Work with backend engineers to integrate with:
• CPU-based AUS simulation on Lambda / similar,
• Model-serving infrastructure.

Risk & Credit Box Design

• Turn model outputs into 
  actionable credit box guidance
  , for example:
• "Tighten here, loosen there — same expected losses, more volume."
• "These borrower profiles are mispriced for their realized performance."
• Run backtests using historical loan performance (e.g., 5+ years of history) to:
• Identify which combinations of FICO/LTV/DTI/income/occupancy performed best or worst,
• Validate that proposed overlays would have reduced losses or improved ROE.
• Collaborate with credit / capital markets to frame:
• Policy changes,
• Pricing adjustments,
• New product specs based on model outputs.

Product & Collaboration

• Partner with:
• Backend engineers (Rust / AWS) to expose models via APIs.
• Product and UX to surface model insights to:
• LOs,
• Underwriters,
• Credit desks,
• Capital markets.
• Help inform how the 
  LLM reasoning layer
   should call into risk models and AUS (tool interfaces, inputs/outputs, error handling, etc.).
• Over time, help recruit and mentor additional ML/DS hires.

What You Bring

Must-Haves

• 5+ years of hands-on experience in 
  machine learning / applied science
  .
• Strong in 
  Python
   and 
  SQL
  , with a track record of:
• Building production-grade models,
• Shipping models into real systems, not just notebooks.
• Deep experience with 
  structured/tabular modeling
  , including:
• XGBoost / LightGBM / CatBoost or equivalent,
• Feature engineering for high-cardinality tabular data,
• Handling imbalanced outcomes and rare events.
• Solid grounding in 
  evaluation
  :
• ROC/AUC, PR curves, calibration,
• Out-of-time and out-of-sample testing,
• Regime / cohort analysis (e.g., by vintage, product, market).
• Experience in 
  one of
  :
• Credit risk modeling (PD/LGD/prepay),
• Insurance underwriting,
• Fraud / financial crime,
• Or a similarly regulated, high-stakes domain.
• Comfort being the 
  first ML hire
  :
• Can design the modeling approach,
• Can work with imperfect data,
• Can document decisions so others can follow later.

Nice-to-Haves

• Direct mortgage / consumer credit experience:
• FICO (especially 10T), VantageScore, agency vs. non-QM, DSCR, etc.
• Experience with:
• Survival analysis / hazard models,
• Time-to-event modeling for prepay/default.
• Familiarity with:
• AWS (Lambda, S3, ECS, etc.),
• ML pipelines (MLflow, Kubeflow, or homegrown equivalents).
• Exposure to LLM-powered products:
• Understanding of how an LLM and a risk model can work together via tools/agents.

Why This Role Is Interesting

• Real moat:
   You're not asking a model to "read PDFs and guess." You'll be training on 
  deterministic AUS outputs
   plus actual loan performance.
• High leverage:
   Your work directly shapes how lenders design and price their credit box—across NQM, DSCR, conventional, HELOC, and more.
• Foundational seat:
   This is a chance to be the 
  architect of the ML stack
   at a company that already has deep lender adoption and a powerful rules engine.
• End-to-end ownership:
   From data design to model to deployment, you'll own a large, meaningful surface area.

Compensation

• Competitive salary and meaningful equity, commensurate with experience.
• Full benefits package.
• Competitive equity options package