Groupe Réussite: edtech marketplace, AI ranking and real-time messaging.

Groupe Reussite. Edtech marketplace matching students, parents and private tutors across 60+ scholastic and extra-scholastic subjects. Product co-build with Abbeal Studio (2019-2023). A platform still in production.

Context

Aghilas Hached, Jonathan Molon, Remy Beharaing and Slim Ghorbal founded Groupe Reussite with one obsession: making student / private tutor matching as simple as an Airbnb booking. Not a directory, not a phone call: a fluid, dual-interface, transactional marketplace.

The challenge

• Build a scalable edtech marketplace with dual interfaces (students/parents vs tutors)
• Handle matching, booking, secure messaging, payments, pedagogical follow-up
• Hold the load on September peaks (back-to-school) without degrading experience
• Let tutors manage their classes, schedule and payments from a dedicated space
• Ship a 'Tutor Space' (lightweight LMS) module in four months without disrupting the main platform

Abbeal Studio execution (2019-2023)

Studio engagement on fixed-price batches with a pilot team led by Adrien Casanova (Abbeal CTO), with Aurelie Largent as senior developer, then Nikhil Kohli and Baptiste Manach. Working mode: sprints, ticket-level estimates, transparency on quotes, progressive skills transfer to the client's internal team.

• MVP scoped 2018-2019, project 'Forfait Reussite' kickoff in November 2019
• V1 launched 2019-2020, first live version of the platform
• Iterative batches 1.1 / 1.2 / 1.5 / 1.5.1 / 1.6 over 2021-2022: functional ramp-up
• Batch 1.6.5 (August 2022): tech debt refactor, Next.js 12.2 migration, bundle optimization
• 'Tutor Space' (Aug-Dec 2022): new module dedicated to tutors (lightweight LMS)
• Maintenance and support 2023 then transition to internal team

The stack we ship today

On the same business (multi-role edtech marketplace, multi-criteria search, tutor ranking, real-time messaging, payments), here's what we assemble today:

• Next.js 16 + React 19 on Vercel: Server Components by default, ISR on tutor profiles, edge fetch of availabilities
• Algolia as search engine: subject x level x location x format, sub-100ms edge latency
• Mistral 7B embeddings + pgvector for semantic student / tutor matching: 'my 10th-grader needs help in physics-chemistry, ideally caring, Tuesday evenings' -> relevant list
• Multi-criteria ranking algorithm: hours taught + reviews + responsiveness + regularity + platform interactions, with optional LLM re-ranking
• Pusher for real-time student / tutor messaging with presence and read receipts
• Stripe Connect for split payments and deferred post-class payments
• Snowflake + dbt data pipeline for internal BI (conversion rate, tutor churn, student NPS)

Why it's hard

• Extreme seasonality: September peak 10x average traffic, capacity ramp-up to anticipate
• Dual interface to maintain in parallel without cross-debt (students/parents vs tutors)
• GDPR compliance on minors: parental consent capture, right to erasure on pedagogical assessments
• Matching quality: bad ranking and 30% of first contacts don't convert to lessons
• Legacy / new compatibility: years of platform with progressive migrations, no big bang

What this engagement brought to Abbeal

Groupe Reussite was our first major edtech engagement. We refined our fixed-price by batches method with ticket-level estimates, learned to ship during September peaks without breaking, and above all trained a generation of Abbeal devs on React + Next.js in a real and demanding product context. That's the experience we replay today at other multi-role marketplaces (real estate, health, pro services).