Portfolio
- Shared definitions and validation so downstream jobs read the same data contracts
- Batch workflows with deterministic outputs, versioned artifacts, and clear failure paths
- APIs, monitoring, and review artifacts that make data products easier to operate
I care about data systems that stay understandable after the first run: where definitions live, how inputs are validated, what gets written out, and how another person can audit the result. These repositories focus on pipelines, API boundaries, batch automation, data quality, and reproducible artifacts. Smaller interactive utilities sit on Data tools.
GitHub profile — opens everything public on my account. On this page, each project card links to that repo and its live demo in the footer.
Projects in this track
Flagship
Feature store (mini)
Built a deterministic feature-table workflow with one definitions registry, schema checks, and repeatable batch builds.
Batch compile turns raw extracts into a versioned feature table under locked column specs and validation rules. Training, scoring, and review paths all read the same built artifact instead of duplicating transform logic.
System skill: Shared data definitions, schema validation, repeatable batch compilation
Review focus: Inspect the definitions registry, validation checks, CLI build path, API transform endpoint, and tests.
Boundary: Not an enterprise feature platform; intentionally scoped to a small, reviewable feature-table workflow.
Flagship
Batch scoring pipeline
Built a scheduled data workflow with aligned preprocessing, deterministic row outputs, and CLI-friendly failure handling.
Batch job processes CSV rows with a fixed preprocessing path and writes score, label, model version, and timestamp on each row. Built for scheduled runs with audit-friendly outputs and explicit failure exits.
System skill: Scheduled file processing, deterministic outputs, run metadata, failure handling
Review focus: Inspect CLI behavior, preprocessing path, output schema, manifest handling, and failure exits.
Boundary: Scoring is batch-first; the live page is an I/O showcase, not the runtime job scheduler.
Flagship
Data quality & monitoring pipeline
Built a batch monitoring workflow for schema, drift, and prediction-quality checks with generated review artifacts.
Validates each incoming batch against a fixed reference and surfaces shifts in inputs, categories, and prediction patterns. Emits batch-level reports and structured artifacts so changes can be reviewed without rerunning ad hoc notebooks.
System skill: Schema checks, drift detection, batch-level reporting, review artifacts
Review focus: Inspect the baseline/current batch comparison, generated HTML/JSON artifacts, thresholds, and test coverage.
Boundary: Covers batch-file monitoring only; no streaming, Kafka, per-request logging, or retraining loop.
Supporting
RAG document data API
Built a document ingestion, retrieval, and FastAPI serving path with citations, source metadata, and deployment boundaries.
Ingests chunked documents into a searchable retrieval index, then serves answers through an API using retrieved context. Each reply cites document, page, and chunk identifiers so outputs stay traceable to source data.
System skill: Document ingestion, retrieval indexing, API serving, source traceability
Review focus: Inspect chunk metadata, retrieval thresholds, API health/serving boundaries, and citation behavior.
Boundary: Useful supporting system, but not the lead Data Systems proof because it carries more AI-specific framing.
Supporting
Churn scoring API
Built a small scoring API around a pinned artifact, request contract, threshold logic, and reproducible serving boundary.
Serves per-account score, tier, and routing flag from one pinned artifact and cutoff. The API keeps notebook workflows off the request path and makes the input/output contract easy to inspect.
System skill: API request contract, artifact loading, threshold logic, serving boundary
Review focus: Inspect the FastAPI request/response shape, pinned artifact path, threshold handling, and stated scope limits.
Boundary: Supporting API example on a teaching dataset; not positioned as a production data platform.
This page is the portfolio index for these five projects: each card carries its GitHub repository and live-demo URL in one place. I update links when a repo moves or a deploy URL changes.