Jan 2026 · LLM System · RAG · Assignment Evaluation
DefendersConnect
An LLM tutor for a cybersecurity education business
100K+
Courses sold through platform
Nov→Jan
2025 → 2026 build window
Curriculum-bound
Constrained answer surface
A domain-constrained LLM system that tutors students, generates and evaluates assignments, and produces course material — so teaching capacity grows with enrolment.
Client
DefendersConnect, an online cybersecurity course business.
The ask
Scale teaching without scaling headcount. Students working through security material need answers, assignments and feedback at a volume no instructor team can serve.
What discovery found
Course completion in self-paced security education depends on how quickly a stuck student gets unstuck. Static content plus a support queue loses people at exactly the point where they'd have learned the most.
What we recommended
A domain-specific LLM system rather than a generic chatbot bolted onto the LMS — one that tutors against their actual curriculum, generates and evaluates assignments, and produces course content, so the teaching capacity grows with enrolment instead of against it.
What we built
A specialised LLM covering tutoring, assignment handling and content generation, deployed into their course platform. November 2025 to end of January 2026.
The hardest call
In security education, a confidently wrong answer is worse than no answer — a student who learns an incorrect practice carries it into real systems. We constrained the system to the curriculum and their verified material rather than letting it answer freely, trading conversational range for correctness.
Results
The system is in active use across their student base, and Zyphex remains their technical partner.
Next project
Health EAI
Clinical Decision Support · Longitudinal Records · SaaS