We built the tutor that works on exactly the right problem.
Founded in Raleigh, NC in 2024. Not a homework helper, not a full-curriculum platform. A single focused tool: watch where a student stalls, generate the next three questions at that exact skill edge.
The pattern that kept showing up
Selena Ortiz spent years working with K-12 students in Raleigh as a learning specialist — in tutoring sessions, study groups, and test-prep coaching. One pattern kept surfacing: after a missed test, students would go back to page one of the chapter. Not because they'd forgotten everything. Because neither they nor their tutor knew precisely what to focus on.
"A student would spend 90 minutes on a chapter they mostly understood, just to practice the 15 minutes of material they actually needed," she recalls. "The issue wasn't work ethic or time. It was resolution. We couldn't see the gap precisely enough to practice it precisely."
In 2024, Selena brought in a curriculum designer and an ML engineer and started building the tool she'd always wanted: something that reads a student's problem-set behavior in real time, identifies the specific skill node where understanding breaks down, and generates three practice questions aimed at exactly that point. Not a chapter. Not a hint system. Three questions at the exact edge.
Tutorwren is not a homework-answer service, not a full-curriculum platform, and not a replacement for teachers. It's a targeted practice tool for the moments when a student knows something is off but can't tell what.
Small team, clear purpose
Educators and engineers who've seen what targeted practice can do.
Learning specialist and educator. Spent years in K-12 tutoring in Raleigh before co-founding Tutorwren to solve the problem of imprecise study practice.
Curriculum designer with a background in K-12 math and science instruction. Defines the skill-node taxonomies that stall detection and question generation operate against.
Builds and maintains the stall-detection and adaptive question-generation models. Previously worked on ML inference pipelines before joining Tutorwren at founding.
Three values that guide every decision
More practice isn't better practice. Three targeted questions at your exact edge move you further than 50 random ones. We optimize for precision, not quantity.
Finishing a chapter without closing its gaps creates false confidence — harder to fix later than if the gap had stayed visible. Tutorwren's session feedback shows where understanding held and where it didn't, not just whether a session was completed.
Our students include minors. We design with COPPA in mind: no advertising targeting, no selling of personal data, parental consent for accounts under 13. Practice session data is used only to generate practice for that student — nothing else.