Industry

Education

Learning systems that respect students and teachers

I build edtech platforms that hold up to the realities of K-12 districts, higher-ed IT, and lifelong learning, accessibility, student data privacy, and the kind of reliability that survives the first day of school.

At a glance
Regulations
5 frameworks
KPIs tracked
6 core metrics
Reference stacks
5 patterns
Services
4 engagements
Case studies
2 published
Perspective

How I think about education

The architecture, the trade-offs, and where I push back on conventional wisdom.

Education is the industry where the user is rarely the buyer, and the buyer is rarely the regulator who can shut you down. That asymmetry shapes every architectural decision. I build edtech platforms that win on three fronts simultaneously: students and teachers actually want to use them, district IT can deploy them without a six-month security review, and the privacy posture survives a state attorney general's questionnaire. My approach starts with rostering and identity, because that's where most edtech projects either succeed quietly or fail loudly.

LTI 1.3, OneRoster, Clever, and ClassLink are the four integrations that decide whether your product is deployable at scale. LTI 1.3 with Deep Linking and Names and Roles Provisioning is the floor for any tool that lives inside an LMS like Canvas, Schoology, or Brightspace. Rostering through Clever and ClassLink is what makes single-sign-on work for 50,000-student districts on day one. I build these integrations with a strict separation between authentication, authorization, and rostering, because conflating them is how teams end up with the same student appearing in three classes they shouldn't see.

Privacy in education is not a posture, it's a contract. FERPA gives parents and students rights over education records, COPPA restricts data collection from under-13 users, and state laws like California's SOPIPA and New York's Education Law 2-d impose stricter overlays. I design platforms where PII is minimized at the schema level, retention policies are enforced in code, and any cross-tenant analytics flows through differential-privacy or aggregation gates. The DPA template you sign with a 100,000-student district will be reviewed by a procurement attorney, and it had better match what the system actually does.

Accessibility is where edtech earns or loses its right to operate. WCAG 2.2 AA is the legal floor in most public education contexts, and Section 508 applies to anything sold to federal or federally funded programs. I build with accessible primitives by default (Radix, Reach UI, native semantics), test with real screen-reader users, and treat keyboard navigation and color contrast as launch blockers, not v2 work. Performance matters too: a slow-loading quiz on a Chromebook over school WiFi is an accessibility issue, not a perf issue.

AI in education is the most exciting and the most fraught application of LLMs I work on. Tutoring agents that meet students where they are can be transformative, but they can also confidently teach wrong math or generate culturally insensitive content. I ship AI tutoring with retrieval grounded in vetted curriculum, an evaluation harness graded by educators, and explicit teacher-in-the-loop review for any output that goes into a gradebook. The bar is higher than consumer AI, and it should be. See a recent edtech build or start a project if your platform needs to ship before the next school year.

Challenges

What teams struggle with

The recurring problems I see on education engagements.

  • 1Student data privacy under FERPA, COPPA, and state laws
  • 2Accessibility compliance (WCAG 2.2 AA, Section 508)
  • 3LMS and SIS integration (LTI 1.3, OneRoster, Clever, ClassLink)
  • 4Massive synchronized load on first-day-of-school events
  • 5Procurement cycles that demand security questionnaires before pilots
How I help

Capabilities I bring

Concrete engineering work that resolves the challenges on the left.

  • FERPA and COPPA-aligned platform architecture
  • LTI 1.3 / Deep Linking / Names and Roles integrations
  • OneRoster, Clever, and ClassLink rostering
  • AI tutoring with guardrails and educator-in-the-loop review
  • Accessible UI built to WCAG 2.2 AA and tested with real users
Metrics

What teams measure

The KPIs leadership obsesses over in this sector. Most tie back to performance and architecture decisions made years before the dashboard was built.

01

First-day-of-school uptime

The single most-watched metric in K-12 edtech; minutes of downtime become district-wide news.

02

Time-to-roster

Hours from district contract signed to teachers and students logged in via Clever/ClassLink.

03

Activation rate per cohort

Share of enrolled students who complete the first meaningful learning action; ties directly to renewals.

04

Accessibility audit score

Independent WCAG 2.2 AA pass rate; required for most public-sector procurements.

05

DPA turnaround time

Days from district legal review to signed Data Processing Agreement; often the bottleneck on deals.

06

Engagement per learning hour

Adaptive systems track meaningful interactions per minute, not vanity time-on-task.

Reference stacks

Stacks I see most often

Patterns I reach for first when scoping a educationengagement. I don't pick technologies for novelty - read more about how I choose.

1

Next.js + Postgres on AWS or Vercel, with LTI 1.3 via ltijs and rostering via Clever/ClassLink

2

Django or Rails monolith for content authoring, fronted by React student experiences

3

OpenAI or Anthropic for tutoring, grounded in pgvector retrieval over vetted curriculum

4

Snowflake or BigQuery for learning analytics, with FERPA-aware aggregation gates

5

Auth0, Clever, and Google Workspace for Education for SSO across districts

Technologies

Tools of the trade

The platforms and frameworks I lean on for education work.

Building for Education?

Let's discuss your specific challenges and how technology can help you ship safely, sleep well, and keep regulators happy.