An independent technology studio designing and building at the intersection of applied AI and product.

Areas of Work
  • AI & Machine Learning
  • Ed Tech
  • Health Tech
  • Applied Research

Claudesque

An agentic chatbot starter kit, built so client AI features can clear the bar set by ChatGPT and Claude.

Analyzing Qualitative Data with LLMs

Multistep LLM workflows that analyze free text, log data, conversation, and observational video — methods borrowed from qualitative social science.

Forest Health

Greenfield ML pipelines for identifying trees in satellite imagery, built to help track the spread of Southern Pine Beetles.

Qualio

Interfaces for analyzing LLM-authored qualitative reports — traverse from summary down to row-level source data.

legible

legible

A mobile Twitter client built to help intermediate language learners grow vocabulary through real social media content.

Squidgies

Squidgies

AI-native language practice for intermediate learners — open-ended writing and roleplay that scales because of LLMs.

About

Legible Labs is an independent technology studio led by Bill Roberts. We partner with founders and product teams on applied AI, machine learning, and the harder edges of getting research-grade work into shipped products.

Recent work spans agentic chat interfaces, multistep LLM workflows for qualitative analysis, ML pipelines for satellite imagery, and AI-native language learning.

Based in Brooklyn. Get in touch at bill@legiblelabs.com.