All work

APPLIED AI · CO-FOUNDER ENGAGEMENT · PRODUCT STRATEGY

An AI tutor built around how people actually learn.

Voice-first, screen-aware, memory-rich. Engineered over months of commercial, product, design and AI work.

OLi logo

CLIENT

Olympus Insights

YEAR

2025

SECTOR

Education technology

SCOPE

Product strategy · UX/UI design · Web app development · Hosting & DevOps · Ongoing support · AI

PLATFORM

Web

Book a call

SHARE THIS

THE BRIEF

Building an AI tutor that actually teaches.

Most AI products in education are LLM wrappers with a system prompt. We started from a different premise. Tutoring is a discipline with its own craft: memory, difficulty calibration, Socratic questioning, multi-modal understanding, emotional awareness. An AI tutor worth the name has to be engineered around that craft, not draped over a generic chat interface.

Over the past several months we have been building OLi alongside Olympus Insights as technical co-founders. Commercial strategy, product strategy, design and AI engineering.

THE CHALLENGE

A loud category, and nothing built for teaching.

The AI tutor category is loud and getting louder. Consumer chatbots are demo-polished but do not sustain learning. Enterprise LMS platforms are bolting generative AI onto legacy infrastructure. Specialist tutors are carving narrow verticals. None of them were built for tutoring from the ground up.

Olympus Insights had what most founders do not: 145,000 or more learners already on the Olympus Academy platform, and a sharp conviction about where education is heading. What they needed were technical co-founders who could build the product itself, a commercially defensible and pedagogically rigorous AI tutor, rather than integrate someone else's API.

The commercial problem was the sharpest of them: carve a defensible position against consumer AI tutors, enterprise LMS entrants and specialist platforms, with a wedge that scales into enterprise.

OUR APPROACH

Strategy first. Then design. Then build.

1. Commercial strategy

We positioned OLi on tutor depth rather than LMS breadth. Generic AI tutors compete on interface, enterprise LMS platforms compete on feature count. OLi competes on teaching: voice-first, screen-aware, memory-rich. That same thesis carries OLi for Business into enterprise conversations without trying to out-LMS the LMS category. Pricing, packaging and go-to-market were built around that wedge, with a living competitive landscape and the consumer wave.

2. Product strategy

We structured a 31-epic, 109-story roadmap across four strategic horizons: core tutor differentiation, platform and integration expansion, niche content verticals, and community and credentialing. Sixteen foundational epics sequence the pre-launch runway, covering authentication, onboarding, chat, voice, sessions, billing, admin portal and analytics, so launch does not trip on plumbing. Every epic maps back to either a commercial wedge or a pedagogical primitive.

3. Design

Voice, screen share and adaptive memory are primitives, and every interaction surface is designed around them. Tutors have personality and a teaching style, whether Socratic, example-led, direct or story-led, and they remember a learner across sessions. ADHD-optimised sessions, dyslexia-friendly interface patterns and CEFR-aligned language tracks were designed in from the first sketch. The result feels less like a tool and more like a relationship.

4. AI engineering

We built and trained the AI system that powers the tutor. Pedagogical behaviour is baked into the stack rather than coaxed out of a prompt: Bayesian difficulty calibration, Socratic questioning patterns, spaced repetition scheduling, multi-modal reasoning over screen content and uploaded images, and concept-level memory that persists across sessions. The model learns each learner's pace, style and weak points and tutors accordingly. This was the hard part, and it is what makes OLi hard to copy.

WHAT WE BUILT

A tutor, and the platform around it.

The tutoring behaviour sits in the system itself, not in a prompt. Around it, the platform pieces a commercial product needs.

  • Voice-first tutoring with real-time screen-share coaching
  • Multi-modal reasoning over screen content and uploaded images
  • Concept-level memory that persists across sessions
  • Bayesian difficulty calibration and spaced repetition scheduling
  • Tutor personalities and teaching styles: Socratic, example-led, direct, story-led
  • ADHD-optimised sessions, dyslexia-friendly patterns and CEFR-aligned language tracks
  • Authentication, onboarding, chat, voice and session management
  • Billing, admin portal and analytics built for enterprise buyers

OLi Tutor role play mode: scenario setup with temperament and difficulty settings, and a live voice session with a hostile customer

Role play: pick a scenario, set the temperament and difficulty, then practise the conversation out loud.

THE FILM

See it for yourself.

The OLi Tutor in action.

THE RESULT

Months of work. One platform.

OLi Tutor is now live at oli-tutor.com.

Voice-first AI tutor

A voice-first AI tutor with real-time screen-share coaching, multi-modal input and cross-session memory, production-ready.

Not just an LLM wrapper

An AI system built and trained specifically for tutoring behaviour, rather than a thin wrapper around a general chat model.

Defensible market positioning

Commercial positioning defensible against consumer AI tutors and enterprise LMS entrants, with OLi for Business as the enterprise wedge.

Launched to market

Launched into a base of more than 145,000 existing learners on the Olympus Academy platform, with admin, billing and analytics built for enterprise buyers.

Building something that should exist?

Book a free 30-minute call. We'll talk through what you're working on, what we'd do, and whether we should partner. No pitch deck, no PDF brochure.

Book a free 30-minute call

Book a free 30-minute call

© 2026 Castle Digital. All rights reserved.