What

The tooling thread: shortening the distance between collecting data and knowing what it means. It started inside clinical trials, where that distance is measured in weeks of manual work, and continues in the independent tools, where it has to close in minutes or nobody keeps using them.

At Alto it produced three tools that reached production — automated screening of patient eligibility against protocol criteria, an FDA audit-level documentation workflow, and a speech task battery built on GenAI-generated stimuli. Independently it is the whole shape of the products: a word spoken during a shooting routine becomes a stat, a trend, and a daily report without anyone opening a spreadsheet.

Why it mattered

Most of the cost in physiological research isn’t the analysis. It is everything wrapped around it — checking a candidate against inclusion and exclusion criteria, writing documentation an auditor will read, producing stimuli, transcribing, reconciling, reporting. Those are exactly the tasks language models are good at, and exactly the tasks that decide how fast a study, or a habit, actually moves.

My role

Built and shipped the three Alto tools, working across clinical, engineering, and drug-development teams to define what each one had to be correct about. The independent work is the same idea with no trial around it: LLM agents and GenAI tooling to shorten the path from acquisition to insight.

Outcome

All three tools reached production use. The speech task battery, with its generated stimuli, is running in a live clinical trial. The independent tools ship weekly, and the standard each is held to is whether it survives daily use by the person who built it.