PlanningWatch
31 councils · nightly
31/31 reporting
A live SaaS that tracks every planning application in Ireland. Nightly ingestion, per-council health checks, an API with webhooks, and paying users. The hero chart above is its data.
read more →I'm a data scientist in Dublin. I take questions from raw data to production — ML research at one end, a live SaaS watching planning applications across 31 Irish councils at the other. The chart on the right is real.
available for data science roles
planning applications · last 14 days
live
1254 applications lodged across 31 councils · updated 2026-09-07 09:38 UTC
31/31 councils reporting · rebuilt nightly 04:30 UTC · built 2026-09-07 09:38 UTC
31 councils · nightly
31/31 reporting
A live SaaS that tracks every planning application in Ireland. Nightly ingestion, per-council health checks, an API with webhooks, and paying users. The hero chart above is its data.
read more →24% → ~50% win rate
filter on
An XGBoost confidence filter trained on 3,033 historical signal outcomes. It doubles win rate by dropping low-confidence signals before they reach the executor. The failed strategies are documented too.
read more →5 directions · 9 experiments
2.6× vs uniform
Testing whether thermodynamics, renormalization group flow, and mean-field theory can guide LLM compression. Entropy-based quantization came out 2.6× better than uniform on GPT-2.
read more →ML & data — Python, XGBoost, PyTorch, pandas, PostgreSQL
Engineering — Django, Rust, Docker, TimescaleDB, CI/CD
Product — Stripe, programmatic SEO, analytics, technical writing
If this looks like your kind of colleague, the CV has the full picture.