AI & Machine Learning
Custom predictive models, forecasting systems, and deep-learning pipelines designed around your data and shipped to production.
WATO builds AI, probability, and data models that turn uncertainty into a measurable edge — for startups, operators, and research teams who need to make the right call under pressure.
Four disciplines, one goal: quantify the uncertainty in your business and build the systems that act on it.
Custom predictive models, forecasting systems, and deep-learning pipelines designed around your data and shipped to production.
Monte Carlo simulations, Bayesian inference, and scenario models that make risk legible instead of intuited.
Decision engines and probability-based algorithms tuned for real workloads — latency, throughput, and interpretability.
Dashboards, APIs, and internal tools that turn a model into something your team will actually use every day.
We take on problems where the right model is the difference between a good quarter and a bad one.
Cut overruns and delays before the first pour. We build models that forecast cost, schedule, and material risk.
Better care, lower waste. We help hospitals forecast demand, catch fraud, and deploy capacity where it matters.
AI and blockchain tooling for valuation, transactions, and underwriting.
Cut delivery cost, reduce stockouts, and keep fleets moving.
We integrate blockchain where it actually earns its keep — transparency, provenance, and settlement.
Statistical models for traders, funds, and treasuries that need to measure risk, not guess it.
Reduce waste, tighten margins, and price smarter with models built for real operators.
A sample of projects we've shipped.
An ML-driven signal engine for forex and equities, built for real-time decisioning on noisy market data.
A model that reads free-text clinical notes to surface likely conditions, improving triage accuracy for care teams.
An image and data pipeline for cancer tumor boards. A new case is matched against the archive of past cases to surface the most similar ones and the decisions those boards made.
A high-performance query layer over large astronomical datasets, cutting analysis time on space science workloads.
Probability-based risk frameworks used by investment teams to size positions and stress-test strategies.
Tools we built and run ourselves. Open any of them and look around.
Small jobs near you, posted by neighbors. Drop a task in when you need a hand, or pull one out and earn. It works for both sides.
Open TaskJarA travel guide for any city. Type a place and see where to eat, what to visit, parks, gyms and day trips nearby, then bookmark your picks and share what you found. On the web and as an iPhone app.
Get it on the App Store Open on the webDrop in a bank or card statement and see where the money went: by category, by month, business versus personal. It runs in your browser and nothing is uploaded. Includes a paycheck plan for people starting to invest.
Open GastosPapers, reports and open tools behind the consulting work. Everything here can be read or run.
An independent replication of a published intraday momentum strategy on 1-minute SPY data. The in-sample result reproduces almost exactly (Sharpe 1.34 over 2015 to 2024), but on the two years that postdate the original study the edge decays: Sharpe 0.39, a 9.4% return against 29.5% for buy-and-hold, and a statistically significant drop in risk-adjusted returns.
Paz, S. (2026). Out-of-Sample Evaluation of an Intraday Momentum Strategy for the S&P 500 ETF (SPY). SSRN Working Paper. doi:10.2139/ssrn.7290621
Read on SSRN Download the PDF Code and dataNLP research that extracts the causes and symptoms of Hashimoto's disease from medical text with BERT and topic modelling. Summer 2021 REU, selected for presentation at ERN 2022.
Read the reportA clinical BERT model trained to find and classify deep vein thrombosis and pulmonary embolism in medical histories, used to study their incidence in cancer patients undergoing endoscopic procedures and any link to general anesthesia.
An analysis tool for the TESS Input Catalog: an interactive sky map with selectable sectors, crossmatching against the catalog through Vizier, linked HR diagrams and histograms for filtering, and CSV export of target lists up to a thousand stars. Python, Flask and the MAST archive.
See the codeTransformer models that classify social media articles as misinformation or disinformation, built during the Fall 2021 REU program.
See the codeMachine learning models that estimate a person's likelihood of developing diabetes from their medical history. Spring 2022, with scikit-learn and Keras.
Ongoing work on algorithmic trading for index futures: an opening-range breakout and fair value gap strategy for the Nasdaq-100 with a Python backtest and a MetaTrader 5 expert advisor, plus a large backtesting pipeline that runs on HPC clusters.
See the codeWATO is an independent consultancy focused on AI, machine learning, and applied probability. We work with founders, operators, and research teams who are tired of dashboards that describe the past and want systems that help them act on what's next. Every engagement starts with one question — what decision is this model supposed to improve? — and works back from there.
Tell us the decision you're trying to get right. We'll tell you whether we can help — and what it would take.