Skip to content

Portfolio · Education technology

A question that cannot waituntil the next lesson.

A student stuck on a problem the night before an exam does not want a booking form; they want somebody now. The marketplace connects them to a tutor on demand across every subject, and holds three roles on the same record — the student who asks, the parent who pays and watches, the tutor who answers — each seeing the part of a session that belongs to them. On top of it sits a recommendation engine that learns which tutor to put in front of whom.

The product is not a lesson. It is the minute between a student being stuck and somebody answering.
Sector
Education technology — on-demand tutoring
Roles
Three — student, parent, tutor — on one record
Matching
On demand, by subject, with learned recommendations
Platform
Serverless: managed identity, storage, functions and recommendations
01

The ask

/ 04

The product is the minute between stuck and answered.

Scheduled tutoring answers a question a week after it was asked, by which point the student has either worked around it or given up on the subject. Here the unit is the doubt: a student raises one — a homework step they cannot follow, a concept that will be on the paper on Friday — and is connected to a tutor who can take it. Everything else in the product exists to keep that minute short.

  1. 01

    A doubt, not a booking

    The unit is one question, which is how students actually get stuck.

  2. 02

    Exam week is the test

    The product is judged on the night before a paper, not on a quiet Tuesday.

  3. 03

    Subject before availability

    A free tutor who cannot take the question is not a match, however fast the match was.

One doubt, asked in the student’s own words, is the unit the whole product is built around.
Measured in minutes, because a question answered next week is a subject already abandoned.
By subject first, then by who is free — in that order, because the second is useless without the first.
02

Three roles, one record

/ 04

The student asks, the parent pays, the tutor answers.

Three roles with genuinely different rights over the same session is the part that is easy to get wrong. A student needs to reach a tutor without an intermediary. A parent needs to see that it happened, and what it cost, without sitting inside the conversation. A tutor needs their availability, their sessions and their earnings, and nothing about anybody else’s. Each role is a view of the same record rather than three products sharing a database.

  1. 01

    A view, not a copy

    One session record, read three ways, so nothing has to be kept in step between apps.

  2. 02

    A parent watches, not listens

    Oversight without being inside the conversation is the distinction the design turns on.

  3. 03

    Managed identity

    Sign-in, roles and tokens handled by a managed service rather than by the product.

Three roles over one session, each seeing exactly the part of it that is theirs.
Identity is managed rather than hand-rolled — three roles is exactly where hand-rolled auth fails.
What a parent can see is a rule, not a screen somebody remembered to hide.
03

The match

/ 04

Which tutor, learned rather than listed.

A directory sorted by rating puts the same handful of tutors in front of everybody and leaves the rest of the supply idle. A recommendation engine, built on a managed recommendations service and trained on the platform’s own session history, learns which tutor suits which student in which subject — so the match improves as the marketplace runs, and the roster is used across its whole width rather than at its top.

  1. 01

    Learned from sessions

    Trained on the platform’s own history, which is the only data that knows these subjects and these tutors.

  2. 02

    The whole supply, used

    A ranked list concentrates demand on a few tutors and idles the rest of the roster.

  3. 03

    Better as it runs

    Every completed session is training data, so the match is a compounding asset.

The offer is learned from what worked, not sorted by a rating everybody sees the same way.
It improves on the session history it is fed, which is the one asset a marketplace makes for free.
04

A second engagement, the same role

/ 04

And the analytics platform built beside it.

Delivered under the same technical lead, for a separate client: an end-to-end analytics platform. Source systems — SQL Server, Oracle, a managed cloud SQL database, a data lake and plain file systems — are extracted into object storage, catalogued by crawlers and queried through a serverless SQL engine; Spark applications written in PySpark and Spark SQL do the extraction, transformation and aggregation across several file formats; and the result reaches the business as dashboards. It is the same discipline as the marketplace, read from the other end: one path in, and everything downstream reading from it.

  1. 01

    Five sources, one landing

    Relational, cloud, lake and file systems all arrive in the same place before anything transforms them.

  2. 02

    Catalogued, then queried

    Crawlers register what landed, so the query engine has a schema nobody had to maintain by hand.

  3. 03

    It ends in a decision

    Aggregation exists to answer a question about customer usage, not to complete a pipeline.

Five kinds of source system, one landing zone, and one catalogue over the top of it.
Catalogued rather than copied, so a table is found by what it is instead of by who made it.
It ends in usage patterns somebody acts on — the pipeline is the means, not the deliverable.

The last word

This shows how Famysys builds a marketplace and the data platform under it: make the unit the thing the user actually has — one doubt, right now — give every role a view of the same record rather than an app of its own, and learn the match from what has already worked. If your supply is idle at the bottom and overbooked at the top, the ranking is the problem.

Built in

  • On-demand matching
  • Three roles
  • Managed identity
  • Recommendation engine
  • Serverless backend
  • ETL pipelines
  • Spark and PySpark
  • BI dashboards

All projects

Let's build what's next

Technology alone doesn't transform businesses.The right partnership does.

Modernizing systems, building a new product, or exploring AI-driven transformation — Famysys helps you move forward with confidence.

Schedule Executive SessionEmail Us Directly
  • SOC2 Type II Compliant
  • Strict Commercial NDA
  • Zero Lock-In Guarantee

Famysys

We Engineer Clarity

Scan to Connect

Instant digital business card & WhatsApp link