Skip to content

Repository files navigation

Data Portfolio

Data engineering, applied analytics, and machine-learning portfolio by Tommi Saltiola.

Canonical GitHub repository

The portfolio emphasizes reproducible analysis, explicit data contracts, deterministic synthetic fixtures, browser-runnable evidence, and test-driven delivery. GitHub source is the portfolio front door; Molab derives each interactive Marimo app from that reviewed source on demand.

Flagship projects

Project Run Engineering evidence Tests
Data Engineer Certification Case Study Molab Four-source wellness pipeline, schema and grain contracts, unit normalization, rejected-record ledger, source profiles, deterministic hashes 36
Data Scientist Certification Case Study Molab Group-aware imputation, three-model validation benchmark, recall-constrained decisions, reserved-test uncertainty, calibration and slice evidence 36
Public-sector Opportunity Pipeline Molab Heterogeneous ingestion, deterministic incremental merge, watermarks, retries, Prefect boundary, transparent scoring 66

All three flagships use deterministic fictional data and pass focused tests, Ruff, strict Marimo checks, executed WASM export validation, and Chromium interaction smoke tests.

The wellness and classifier projects are runnable case studies for broad data-engineering and data-science competencies covered by the certifications. They use public source, tests, documented contracts, and deterministic synthetic fixtures. Private assessment materials are neither needed nor published.

Supporting learning labs

These smaller labs modernize five historical learning themes as independently written Marimo apps. They are supporting analytical demonstrations, not claims of production deployment. Every default path uses deterministic synthetic data and runs without credentials, uploads, or private infrastructure.

Lab Run Analytical focus
Airline Delay Quality Lab Molab Flight-grain delay components, cancellation context, and carrier summaries
Synthetic Health Risk Quality Lab Molab Duplicate-profile audit and descriptive ordinal associations over fictional data
Restaurant Location Quality Lab Molab Coordinate validation, accepted records, and an explicit unresolved ledger
Streaming Catalog Explorer Molab Release-period, genre, and duration summaries at title grain
Judo Medal Explorer Molab Medal-rate summaries at declared fictional athlete-event grain

See the Analytics Learning Labs evidence and data provenance.

Professional certifications

DataCamp Data Scientist and Data Engineer certificates are published with official verification links. Each credential points to a runnable public case study, its source, and its engineering specification.

Review certifications and competency mapping.

Certificate images do not admit any associated assessment prompts, datasets, solutions, schemas, metrics, outputs, or grader rules into this repository.

Search and knowledge products

Two separate repositories turn deeper search and knowledge-system patterns into public demonstrations using independently implemented source:

  • Search Taxonomy Lab — TF-IDF and latent-semantic evidence, cluster discovery, transparent classification, human review, benchmarks, and a hash-chained audit ledger.
  • Content Evidence Workbench — retrieval, exact citations, declared-entity context, judged evaluation, and explicit human review over a synthetic corpus.

Fixed-scope service

The DBSCTR Delivery Accelerator adapts an auditable agentic engineering lifecycle to one repository and delivers one bounded pilot through every applicable gate.

Discuss a pilot or inspect the owner-authored, MIT-licensed DBSCTR source.

Reproduce the evidence

Each flagship has its own pyproject.toml and uv.lock. From the repository root, run the matching app path inside each project:

(cd projects/wellness-data-pipeline && uv sync --locked && uv run --frozen pytest -q && uv run --frozen marimo check --strict app.py)
(cd projects/content-performance-classifier && uv sync --locked && uv run --frozen pytest -q && uv run --frozen marimo check --strict src/app.py)
(cd projects/public-sector-opportunity-pipeline && uv sync --locked && uv run --frozen pytest -q && uv run --frozen marimo check --strict app.py)

Project-local README files list their complete Ruff and verification commands. The five learning labs share one locked environment:

(
  cd projects/analytics-learning-labs
  uv sync --locked
  uv run --frozen pytest -q
  uv run --frozen ruff check .
  uv run --frozen ruff format --check .
  uv run --frozen marimo check --strict apps/*.py
)

The quality workflow runs project gates, builds all eight apps in temporary WASM test surfaces, exercises them in Chromium, scans critical vulnerabilities, and emits an SPDX software bill of materials.

Trust boundary

This repository contains no employer code, client data, private credentials, restricted assessment prompts, datasets, solutions or outputs, proprietary SaaS source, or personal datasets. Visitor uploads remain runtime-only and are never admitted to committed session previews.

See DATA_PROVENANCE.md, and DEPENDENCIES.md.

Publication model

Saltiola7/data-portfolio is the canonical public portfolio. Separate static hosting is retired. Molab derives each runnable app directly from reviewed GitHub source. Changes integrate through pull-request review.

License

Owner-authored code and documentation are available under the MIT License. Professional credential images are published solely as verification evidence and are not licensed for reuse under the repository's MIT license. Any future public datasets retain their separately recorded source licenses.

Releases

Packages

Used by

Contributors

Languages