Description

Job Description:

True must-haves:

• Strong, recent, hands-on Python coding
• SQL and Snowflake
• Image processing or computer-vision output experience
• OpenCV or Pillow; PyTorch/TensorFlow exposure is relevant
• JSON and semi-structured data
• Data modeling and data engineering
• Validation of extracted images, attributes, and metadata against source documents
• Root-cause analysis and ability to fix extraction, metadata, and output issues
• Mortgage, appraisal, collateral, or closely related financial-domain experience
• Ability to work onsite in McLean for a short-term contract

Important clarification:

The resource will not primarily build or train computer-vision models. The person will:

• Prepare model inputs and process model outputs
• Extract property images from appraisal PDFs
• Generate and validate attributes and metadata
• Associate images with the correct property and loan
• Load or support data in AWS S3, Snowflake, and a vector database
• Compare extracted information with original source documents
• Investigate discrepancies and determine root causes
• Fix extraction, metadata, and downstream-data problems
• Partner with modelers when the problem is inside the model
• Work mainly with data-engineering and modeling teams

Notebook experimentation and model-training experience are advantageous but secondary.

Best candidate profile:

• Production-level Python
• OpenCV/image extraction
• SQL, Snowflake, JSON, and AWS S3.
• Model-output validation and reconciliation.
• Mortgage/appraisal data knowledge.
• Fannie Mae or Freddie Mac experience.
• Some AI or computer-vision model familiarity.

Strong expertise in Python, SQL, Image analytics, data modeling, and snowflake. Financial / mortgage background is required.

Position Overview:

• We are seeking a hands-on Data Scientist with strong Python and Computer Vision skills to design, build, test, and operationalize capabilities that convert image-based content and model outputs into usable, reviewable, and analytics-ready data products. This role requires strong software engineering fundamentals, computer vision coding experience, data engineering skills, and the ability to partner across product, modeling, engineering, research, and business teams. The Data Scientist will contribute to capabilities for image extraction, metadata generation, model-output validation, quality review enablement, and downstream structured data integration. The role is expected to balance Python development, testing, analytical troubleshooting, and delivery execution to help users review, validate, and act on computer vision outputs.

Key Responsibilities:

 Computer Vision Development & Model Output Engineering

• Develop, enhance, and maintain code that supports computer vision model output processing, image extraction, metadata generation, and validation workflows
• Work with image-based model outputs, bounding boxes, labels, confidence scores, extracted attributes, and structured metadata to support downstream review and analysis
• Build reusable utilities for parsing, transforming, validating, and comparing computer vision outputs across model versions and production-style runs
• Apply strong Python coding practices to automate testing, issue detection, data preparation, and model-output quality checks

Product & QC Workflow Enablement

• Support product capabilities that allow users to review, validate, correct, and quality check model-generated outputs
• Translate computer vision models output into user-facing review patterns, QC screens, exception workflows, and validation experiences
• Partner with UI developers, product owners, and business users to define practical capabilities for model-output inspection and operational review

Test Data, Validation & Quality Engineering

• Create and manage representative test datasets for image extraction, metadata validation, regression testing, and model performance review
• Perform structured testing of model runs across historical and current datasets to identify extraction gaps, metadata issues, formatting errors, and quality concerns
• Validate extracted images, image classifications, and metadata against original PDFs, appraisal reports, and other authoritative source documents to confirm completeness, accuracy, and traceability
• Document defects with clear evidence, expected results, actual results.
  Retest remediated issues and contribute to repeatable quality gates for model-output readiness
• Data Integration, JSON Engineering & Analytics Readiness Develop scripts and data pipelines that convert model outputs into structured and semi-structured formats suitable for research, analytics, and downstream consumption
• Support loading and validation of model outputs as JSON Variant or similar semi-structured data formats
• Ensure extracted image attributes, metadata, and model-output payloads are traceable, consistent, and accessible for analysis

Cross-Functional Delivery & Technical Problem Solving

• Collaborate across product management, model development, UI engineering, data engineering, research, business, and delivery teams to operationalize computer vision capabilities within data-driven products
• Investigate technical issues across image inputs, model outputs, metadata payloads, data loads, and user-facing QC workflows
• Communicate progress, risks, blockers, and technical findings clearly to engineering partners and business stakeholders

 Required Qualifications

• Computer Vision Coding & Software Engineering Hands-on experience coding computer vision or image-processing solutions using Python and common libraries such as OpenCV, Pillow, PyTorch, TensorFlow, or similar frameworks
• Strong ability to process image files, extracted labels, model predictions, confidence scores, annotations, bounding boxes, and metadata payloads
• Experience writing modular, maintainable code for automation, validation, transformation, testing, and troubleshooting

Key Skills

Education

Any Graduate