01
implementedpipeline::ingest
Python lands Garmin source payloads.
Ingestion jobs authenticate to Garmin Connect, download FIT activity files, and write source files to the raw S3 landing zone before modeling.
These same runs sync to Strava.
inspect running_signals
Running data, modeled into signals.
Running Signals is an end-to-end lakehouse project that makes training patterns inspectable—from recoverable Garmin payloads to tested analytical marts and focused data explorers.
Review the project source code to inspect the pipeline, transformations, tests, and interface.
01 / Signal families
Three questions, answered with inspectable data.
The experience stays descriptive. It surfaces evidence and definitions without turning observations into coaching or medical claims.
02 / Methodology
Follow every signal back to source.
Ingestion, bronze preservation, silver standardization, and gold analytical logic remain separate so lineage and quality are straightforward to inspect.
02
implementedpipeline::bronze
Bronze preserves source-shaped evidence.
Databricks parses source files into lineage-preserving bronze tables. S3 retains recoverable source files for replay.
03
implementedpipeline::silver
Silver standardizes reusable entities.
dbt silver models clean, type, deduplicate, and standardize run, record, date, and week entities into tested building blocks for downstream analytics.
04
implementedpipeline::gold
Gold publishes analytical signal marts.
Gold models define consistency, volume, route, and descriptive fitness marts. Selected FIT read models publish to Supabase.
03 / System stack
An inspectable analytics workflow.
Each tool has one explicit job. The architecture favors recoverability, readable transformations, testable definitions, and a deliberately lightweight presentation layer.
eight connected capabilities
Terraform
Bootstraps governed cloud data foundation.
Garmin
Supplies FIT activity and route telemetry.
Python
Automates ingestion into the raw landing layer.
AWS S3
Preserves recoverable raw FIT activity files.
Databricks
Parses FIT files and hosts lakehouse models.
dbt / SQL
Tests and transforms silver and gold models.
PostgreSQL
Serves curated data for downstream applications.
Next.js
Builds and renders the presentation application.
04 / Explore
Explore published running data.
Move from sessions and route geometry to consistency, volume, fitness, and the planned read-only agent interface.
tip: press cmd/ctrl + k