Original housing-market forecasts and applied machine-learning studies, presented with transparent methods, measured limitations, and production context.
Research preview · Week 2
Cambridge leads near term. Ithaca leads since 2020.
The same 80-month protocol produces a split result: Cambridge carries the higher forecast value and modest month-ahead momentum, while Ithaca has delivered stronger cumulative appreciation.
Cambridge 02139 vs Ithaca 14850 · Condo/co-op series · 44 rolling backtests per city
Ithaca growth since 2020+29.5%Observed cumulative change
Applied machine-learning projects
Real data. Reproducible evaluation. Operational context.
Project 01 · Forecasting
Hyperlocal housing forecast pipeline
A repeatable workflow for ingesting monthly market data, comparing multiple time-series candidates, conducting rolling backtests, and publishing decision-ready forecasts.
Python · Time series · Model selection · Explainability
A platform proof of concept that converts experimental workflows into governed, versioned, observable services consumable by scientists and autonomous tools.
MLOps · API contracts · Governance · Observability
Two markets complete. A reusable system is taking shape.
The objective is not simply to publish 64 forecasts. It is to build a comparable evidence base across markets—using consistent evaluation, clear local context, and a production workflow that can scale without sacrificing rigor.
2 / 64 cities published
Next phase: validate the publishing workflow across additional Northeast markets before expanding regionally.
We are building partnerships around local data access, independent model validation, sponsored city research, and applied machine-learning systems. Every engagement starts with a defined question and an evidence plan.
