Origin story
Born from the Lally AI Academy.
Lumber Market Watch was built during the Lally AI Academy — a 30-day experiential sprint at Rensselaer Polytechnic Institute where cross-disciplinary teams design, build, and launch production-ready AI products from scratch.
Starting on Day 1 with only an idea, teams iterate through rapid prototyping and weekly milestones to reach a live deployment by Day 31. Each cohort receives up to a $500 tools budget, access to RPI faculty mentors, and workspace to ship something real — not a mockup.
Teams must include at least one Lally School student and a minimum of three members. All majors and experience levels are welcome — no prior coding experience required. Deliverables include a live deployment, a public code repository, and video reflections documenting the build process. Participants earn a certificate of completion.
The system
Four components, one signal.
Coastal Surge Data
NOAA CO-OPS tide gauges and NDBC buoys stream water-level and wave data every 6 hours across 12+ East Coast stations, capturing the storm events that disrupt timber supply chains.
LSTM Model
A Long Short-Term Memory neural network trained on historical HURDAT2 hurricane tracks and surge observations learns the non-linear relationship between coastal flooding and lumber price spikes.
Macro Filter
Real-time S&P 500 momentum and 10-Year Treasury Yield are checked before any signal is issued. Severe macro regimes can suppress an otherwise valid BUY to protect against false positives.
Automated Execution
Signals are relayed to an Alpaca paper-trading account, recording every entry and exit. The full trade log, P&L, and portfolio history are visible inside the dashboard.
The team