Flashlight builds retrieval systems for personal medical records and whole-genome data, turning overwhelming files into structured, source-grounded information you can understand and discuss with qualified professionals.
Note Flashlight 3D's products are currently in development and do not provide medical advice or diagnosis.
Millions of rows, pages, and raw sequencing lines sit in the dark, scattered across portals, PDFs, and files no one has time to read end to end. Flashlight's systems curate the relevant sources first, then let AI explain what they mean, rather than asking AI to recall an answer from memory.
Medical Records, the Genomic Sequencing Pipeline, and Chippy all move the same way: an unstructured pile is curated into a database, and AI explains what's there instead of guessing.
A whole-genome VCF holds about five million variants. The pipeline curates the ones with available evidence and produces a report that explains their significance in relation to specific genes.
Fragmented PDFs, portal exports, labs, and visit summaries become one longitudinal, source-grounded record you can query and share with your care team.
Built first for Flashlight's own team: an internal agent that answers questions from our working documents, proving out the same architecture at a smaller scale.
A founder, a core team, and a six-person Summer 2026 intern cohort from UC San Diego, Case Western Reserve, and other schools, working across genomic data processing, sequence annotation, AI, application development, user experience, and responsible product design.
Notes on the technical and product problems we're working through, who's working on them, and where things stand.
