Why Tytos Is Different

Built From Experience

I started working in residential construction at 14, spending summers and holidays as a laborer for my father's construction company. What began as a way to earn extra money turned into 9 years of hands-on experience learning firsthand how homes are built, what causes delays, and what separates a project that finishes on time from one that doesn't. By my early twenties I had moved into project management, overseeing crews and timelines and gaining a ground-level understanding of the variables that affect completion.

At 26 I made the move to banking, where I spent the next 12 years managing draws and risk for construction loans — the final 6 of which were dedicated to applying machine learning and data science to risk mitigation. That progression, from assessing risk manually to training models that could surface it automatically, gave me a clear picture of what lenders actually need and what technology could realistically deliver. Tytos is the product of that experience.

Why I Built Tytos

After 24 years working in and around residential construction, one problem kept showing up: construction lenders, developers, and portfolio managers are working harder than ever but the nature of the job means the majority of that effort goes toward finding which projects need attention, leaving little time to actually address them. It's not an effort problem. It's a time problem.

My view is you spend roughly 80% of your time identifying potential problems and only 20% of your time actually addressing them. That balance needs to shift. The identification should be automated allowing more time for the risk mitigation, the relationship work, and the business generation where real value is created.

Tytos was built to do the finding. Deep learning models trained on 138,000+ residential construction projects, with an average margin of error of ±8 days, give you an early warning system that surfaces risk before it becomes a problem so you can spend your time where it actually matters.

The Model

Tytos uses deep learning models trained on real residential construction project data covering all 50 states. The models account for regional differences in typical build timelines, project type, size, and, for higher tiers, actual project performance data to produce predictions accurate to within ±8 days on average.

This isn't a rule-based system or a simple average. It's a model that has learned the patterns across hundreds of thousands of real projects and applies that learning to give you a precise, data-driven answer to the question you really want answered: will this project finish on time?

TJ Songy

Founder, Tytos LLC

Questions or want to talk through how Tytos can work for your portfolio? Reach out directly at tj@tytospredicts.ai