Transit Data Engineer · GIS Developer · Open Source Contributor
Boston, MA
I believe cities move better when their data does too. I spend my days wrangling maps and my evenings chasing transit feeds, writing the pipelines, parsers, and dashboards that turn raw GTFS into something a city can act on.
Professionally, I keep a large ArcGIS Enterprise ecosystem humming: automating the quiet, unglamorous work of account lifecycle management, license reclamation, and organizational reporting so that the people who make maps can focus on making maps.
In my own time, I volunteer with TransitMatters as a software developer, and from May to September 2026 served as interim co-lead of TransitMatters Labs, the org's volunteer engineering program, to help hold the MBTA accountable through data. There's something deeply satisfying about a slow-zone alert that actually reflects a slow zone, a ferry tab that didn't exist last month, or a bus speed map that shows exactly where a route crawls. Small things that riders notice even if they never see the code behind them.
A real-time GTFS-RT ingest and archival pipeline polling 500+ live feed endpoints from 200+ transit agencies across North America, Europe, and Australia/New Zealand.
- Region-local poller hosts (US, Frankfurt, Australia) with continent-based feed routing into centralized, content-addressed S3 landing storage with request-level deduplication; zstd-compressed landing archives halved cross-region transfer
- Cloud-native infra on AWS (EC2, S3, ECS Fargate), provisioned with Terraform: a decoupled poll/rollup architecture with the nightly rollup split into per-stage Fargate tasks right-sized from observed usage, plus Datadog monitoring and cost tracking
- Performance-critical GTFS-RT decoder ported from Python to Rust via PyO3 bindings, alongside agency-specific decoders (e.g. TfNSW, with per-carriage consist data)
- Curated analytics layer (on-time performance, schedule adherence, trip metrics, PostGIS-backed segment speeds) exposed through a FastAPI REST API, including a historic 511 backfill
- Next.js/TypeScript + MapLibre dashboard visualizing real-time delays, service alerts, segment speed maps, and cross-agency comparisons
TransitMatters is a data-driven MBTA advocacy organization.
| Contribution | Details |
|---|---|
| Bus speed map | End-to-end: segment-speed jobs in mbta-performance (container image, Kubernetes CronJobs), bus LAMP ingest, gzipped segment leaderboard, and map/graph views with time-of-day and day-type filters |
| API & client libraries | Rewrote the Data Dashboard OpenAPI spec (real response shapes, operationIds, enums); transitmatters Python client on PyPI with MkDocs docs; CI and spec sync for transitmatters-swift |
| transitmattr | R client package for the Data Dashboard API — API layer generated from the OpenAPI spec, test suite, vignettes, pkgdown site, CI |
| Ferry data | End-to-end: ingestion pipeline, GTFS processing, frontend tab, and stop/route constants |
| Bus route expansion | Silver Line, remaining 3-digit routes, multi-routes (434/435/436), SLW, Fall 2026 Bus Network Redesign updates, and "frequent bus" routes |
| Weather | OpenMeteo weather ingestion and dashboard visualization |
| The RIDE | Paratransit ridership ingestion |
| Added Fields | Full-stack pipe: gobble → mbta-performance → dashboard tooltip |
| Mattapan Trolley | Slow zone support and New Train Tracker integration |
| Internationalization (in review) | react-i18next foundation and UI text extraction across the dashboard (5-PR series) |
| Infrastructure | Test suites for data-ingestion and mbta-performance, docstrings and MkDocs docs, GTFS ingest memory reduction, beta auto-deploy, uv migration, GitHub Actions build offload, Dependabot patches, SharePoint integration, spaCy-based alert classification |
Repositories: t-performance-dash · data-ingestion · mbta-performance · gobble · transitmatters-py · transitmattr · new-train-tracker · regional-rail-explorer
| Project | Description | Stack |
|---|---|---|
| astrotrain | Learned countdown clocks for the MBTA subway: an LSTM trained on LAMP data, scored against the schedule | Python, TensorFlow |
| nibble | GTFS-RT to MBTA V3-style SSE adapter with vehicle-state tracking and stop interpolation | Python, FastAPI, Kubernetes |
| gtfs_rt_rater + frontend | GTFS-RT feed quality rater | Rust, TypeScript |
| anduin | GTFS route map-matching to OpenStreetMap via Valhalla, with shared-segment analysis | Python |
| fmeflow | FME Flow REST API client generated from OpenAPI (434 endpoints) | Python |
| bluebike-trip-explorer + bikeshare-ingestor | Bikeshare trip flow visualization across 5 US systems | TypeScript, Python |
| amtrak-ingestion | Amtrak data pipeline | Python |
| greenline_idletime | MBTA Green Line idle time analysis | Python |
- Fall NEARC 2025 — Northeast Arc Users Group
- Fall NEARC 2023 — Python for ArcGIS User Management



