Data Engineer // Astrophysics Pipeline // Reproducible Systems
// Messy data in. Structured artifacts out.
I build systems that take raw public data through explicit ingest, metadata handling, validation, and analysis-ready outputs.
Across projects, the recurring pattern is staged processing, rerunnable artifacts, and enough provenance to inspect what changed between raw inputs and derived results.
The operating domain here is scientific data, but the transferable value is workflow design, validation discipline, and technical ownership over messy datasets.
Completed work is separated from active systems so finished case studies stay legible and ongoing builds stay honest.
Repositories only move into the curated project set when the workflow, artifacts, and validation story are clear enough to stand on their own.
Active systems stay in a lighter log format so platform direction, tooling, and unfinished work can remain visible without pretending to be complete.
I build Python-centered workflows that take messy records through ingest, metadata shaping, validation, and structured outputs. The datasets are scientific, but the signal is engineering discipline, reproducibility, and traceable execution.
A metadata-aware Python workflow for querying, downloading, indexing, and analyzing public IRIS observation files.
The system keeps raw inputs, per-observation lineage, duplicate-obsid handling, and derived outputs coordinated inside one reproducible pipeline. Archive query, obs_dir-aware storage, metadata indexing, per-observation outputs, and merged event artifacts for IRIS Level 2 data.
A reproducible Python pipeline for ingesting public survey light curves, standardizing them into parquet artifacts, and evaluating period recovery against Gaia DR3 truth data.
Public baseline pipeline for live ingest, standardized time-series outputs, and evaluation under Rubin-like cadence constraints. The system keeps ingest behavior, failure modes, and baseline results visible instead of burying them behind final figures.
A reproducible Python workflow for ingesting, cleaning, binning, validating, and packaging multi-source photometry plus supporting raw-image assets.
Reproducible photometry and raw-image workflow with standardized outputs, overlap validation, and archive-aware provenance. The project keeps analysis products, cross-source checks, and provenance records aligned inside one explicit pipeline.
A reusable Python workflow for clustering, membership inference, quality inspection, and cross-dataset comparison on Gaia-based tables.
CLI-backed clustering and data-quality workflow for Gaia-derived tables, with preserved notebook provenance and multi-cluster outputs. It turns notebook-heavy analysis into a CLI-backed pipeline with clearer reruns and public workflow structure.
Building a metadata-first multi-archive ingestion pipeline for HLSP collections in MAST.
Programmatic access to HLSP data is strong, but production-style ingestion still needs explicit contracts around observation metadata, product manifests, rerun safety, snapshot history, and queryable downstream storage. The repo is structured around three core entities: collections, observations, and products, with bronze raw snapshots, silver normalized parquet, and gold latest views.
A technical-editorial mission archive for Artemis II, built as a longform, single-page experience.
The structure combines mission chronology, public NASA update links, curated physics explainers, official NASA imagery, and explicit source methodology. Built with a local JSON data backbone curated from NASA public sources.
Active builds are logged separately from finished case studies so iteration, architecture decisions, and public-facing system work stay visible without being overstated.
Building Astrolyte as a public-facing data platform where ingest, indexing, processing, validation, and dataset surfacing are treated as explicit stages rather than hidden implementation detail.
The grouping is engineering-first: workflow handling, structured execution, and reproducibility come before the domain context.
Ingest, standardize, validate, and package messy public data into structured artifacts that can be rerun and inspected.
Design multi-stage paths from raw records to derived outputs, with explicit checkpoints, provenance, and visible validation surfaces.
Build domain-specific tools where the scientific context is the operating domain and engineering discipline stays in the foreground.
Recent public repository movement, kept secondary to the curated case-study set but useful as a consistency check.
Astrolyte website and public surface for workflow-oriented scientific data systems.
Personal portfolio for workflow-heavy data engineering and scientific data systems.
Reproducible IRIS Level 2 workflow for archive discovery, metadata indexing, per-OBS quicklooks, and duplicate-window ROI audits.
Reproducible T CrB observational workflow with clean products, overlap validation, and archive-aware support assets.
If you want to talk about data workflows, backend-oriented builds, internships, or collaboration, this is the fastest way to reach me.