Skip to content
Home

Work — 5 studies

Professional case studies.

Data science, machine learning, and data-systems work from my co-op at Command Credit Corp. Each study states what I did, how it was checked, what actually shipped, and where its limits are.

Company work is presented through anonymized summaries and illustrative examples.

Showing 5 of 5 case studies

  • 01 · Data Science / Machine Learning

    Status: Shipped score · separate research

    Commercial credit-risk modeling

    Tested whether commercial credit signals separate businesses with observed adverse outcomes, then built a regularized ranking score that now ships in a business-credit product.

    Technical detail
    • Outcome labels built from public filings with staged name matching
    • Regularized logistic regression on percentile-binned inputs
    • Held-out, temporal, backward, and forward evaluation
    • Python ↔ TypeScript parity on reference examples
    • Python
    • DuckDB
    • NumPy
    • pandas
    • SciPy
    • scikit-learn
    • Jupyter
    • TypeScript
    Read case study

  • 02 · Data Systems

    Status: Shipped matcher · research prototype

    Cross-provider entity resolution

    Matched the same business across inconsistent names and addresses from multiple commercial data providers, with conflict safeguards, a review path for ambiguous cases, and readable decision trails.

    Technical detail
    • Shared normalizer with parsed street components
    • Edit-distance and n-gram similarity with hard conflict rules
    • Link / review / no-link outcomes with decision trails
    • Server-validated user confirmation for ambiguous cases
    • TypeScript
    • React
    • SQL
    • Python
    • String similarity
    • Record linkage
    Read case study

  • 03 · Data Systems / Machine Learning

    Status: Working local prototype

    Public-record business-data pipeline

    A config-driven local pipeline that combines public records into one business reference, with layered processing, provenance, record linkage, and transparent modeling experiments.

    Technical detail
    • Raw → staging → normalized → identity → export layers
    • Deterministic links first, probabilistic linkage for the rest
    • Linker tuned on known-correct identifier pairs
    • Naive Bayes industry classifier with abstention and self-training
    • Python
    • Polars
    • DuckDB
    • Parquet
    • pytest
    • rapidfuzz
    • YAML
    • Census Geocoder
    Read case study

  • 04 · Data Science

    Status: Analysis delivered · displays shipped

    Credit-data quality and benchmarking

    Audited completeness, comparability, and meaning in commercial credit data, and turned the findings into recommendations and more interpretable score displays.

    Technical detail
    • Seeded reservoir sampling from very large files
    • Sentinel-aware fill rates and distribution analysis
    • Rank correlation, weighted agreement, equipercentile equating
    • Midrank percentile displays with deterministic explanations
    • Python
    • pandas
    • NumPy
    • SciPy
    • scikit-learn
    • Matplotlib
    • DuckDB
    • Jupyter
    Read case study

  • 05 · Data Science

    Status: Analysis and recommendations delivered

    Customer and product analytics

    Analyzed usage and purchasing patterns in a legacy business-data product with SQL, to inform the design of its replacement.

    Technical detail
    • Explicit grain with distinct counts
    • Conditional aggregation, time bucketing, cohort comparison
    • Cross-system reconciliation
    • Recurring analyses automated in Python
    • SQL
    • MySQL
    • DBeaver
    • Python
    • pandas
    • Excel
    Read case study