Adaptive Tutoring Copilot
Personalized K-12 tutoring agent grounded in district curricula with assessment-aware feedback.
- Client
- Meridian Education
- Year
- 2025
- Duration
- 12 weeks
- Team
- 3 engineers, 1 designer, 1 curriculum specialist
Meridian serves 32 districts and had watched generic AI tutors confuse students by teaching methods that contradicted the classroom. We built a Socratic tutoring agent scoped strictly to district-approved instructional materials, integrated into the existing LMS, with student privacy constraints designed in from the first architecture review.
The brief
Deliver one-to-one tutoring at scale that reinforces what the teacher taught, withholds answers, and never trains on student work.
- Education
- Tutoring
- RAG
The decisions that mattered.
Every engagement turns on a handful of calls. These are the ones that decided whether this system reached production.
Scoped to the district's own curriculum
Retrieval is restricted to approved instructional materials per district, so explanations match the method students were actually taught rather than a plausible alternative.
Socratic by construction
The agent is constrained to scaffolding and cannot produce a final answer. Persistent gaps escalate to the teacher's dashboard instead of being papered over.
Privacy designed in, not bolted on
FERPA and COPPA-aligned from day one: strict data minimization, no training on student-generated content, and district-level data isolation.
Zero-friction teacher setup
It lives inside the existing LMS with no separate gradebook and no new login. Setup takes minutes, which is the only budget teachers actually have.
What shipped, and what changed.
Outcomes
- 180K students reached across 32 districts
- Mastery gains of 0.4 standard deviations on unit assessments
- 6 teacher hours reclaimed per week on average
- Zero student-data incidents across the first academic year
“The first AI tutor our teachers stopped treating as a threat, because it teaches the way they taught it.”
What we built it with.
- Claude
- pgvector
- Python
- Next.js
- Postgres
- LTI 1.3
Other projects from the shelf.
Clinical Documentation Copilot
Ambient AI scribe that drafts structured notes, ICD-10 codes, and prior-auth letters in real time.
Read moreEquity Research Terminal
Multi-modal RAG over 10 years of filings, transcripts, and broker research with citation-grounded answers.
Read morePredictive Maintenance Platform
Time-series + vision models predicting equipment failure 72 hours in advance across 14 plants.
Read moreTell us about the problem.
A 30-minute call with the senior team. We will tell you what is realistic, what it would cost, and whether we are the right people for it.