Peerz Academy first engaged cloudThing back in late 2018 to build a cloud-based, peer to peer tutoring platform that would enable a student in need of aid with a homework question or revision topic to reach out to qualified and vetted peers for help.

The platform would allow for students to upload images and text describing their question to the platform and then receive 1-2-1 help from a qualified mentor.

Peerz Academy is a secure, online learning platform funded by the NCFE (National Council for Further Education) that allows verified mentors to deliver lesson content created by expert teachers as well as peer to peer help with specific questions or revision topics.

Their aim is to provide an effective and affordable alternative to expensive private tuition by combining the latest in Cloud Based and AI technology with high impact educational strategies.

Their mission statement is to transform learning by enabling peer to peer interaction in a safe and secure environment, enabling effective learning between peers.

Once the Peerz Academy had engaged cloudThing we kicked off their project, as we do with all our customers, with a discovery phase led by one of our Solution Architects involving all key stakeholders.

The aim of the Discovery Process was to fully capture their requirements for the project, get a full scope of possible pain points and to understand the functions and operations the platform would need so we could successfully road map their ambitions within our proposed solution.  

The goal was to build a peer-to-peer learning platform that would transform learning and disrupt the private tuition industry. 

It needed to be a holistic learning app that would allow users to manage revision and seek help from verified mentors on a secure platform.

The app also needed to be monetised using a secure payment model similar to phone networks (pay as you go/subscription).

In its current form the Peerz Academy App was built on a hybrid system using both Microsoft Azure and Google Firebase which would allow for instant updates from the database when needed as well as easy to implement improvements in the future.

Originally specced out as wire frames, we progressed this to higher fidelity designs during the build process using React Native so Peerz Acadamey could more easily see all the individual components of the platform as and when they were created allowing for a much quicker feedback and improvement process.

The finished app has three different functions available to students with a back-end system for the vetting of mentors and processing of all app functions. 

• Peerz Academy Back-End Functionality – Private tutors, teachers and other students can all sign up to the platform to become a mentor.
  Before signing up they need to agree to a vetting process that involves a DBS check (Disclosure and Barring Service) to keep the users of the app safe. Once signed up the mentors then need to upload appropriate certification for the subjects they wish to mentor (and answer questions) on.
  For instance, an English Literature A-Level student could upload their English Literature GCSE certificate which would allow them to answer GCSE level questions whilst a private tutor with an English Literature degree could upload their degree certificate which would allow them to answer GCSE and A-Level English literature questions.
• Answering Specific Questions – The first function of the app allows a student to upload an image of the question they need answering (for a small fee which is split between the mentor and the Peerz academy).
  The app then categorises this by subject and topic (i.e. mathematics then algebra) then level (i.e. GCSE/A-level).
  Once categorised it will be sent to ten mentors ranked on a five-star system (based feedback from previous students) to answer. The first mentor to select the answer will be assigned to it with the student receiving a notification that a mentor is working on their answer.
  If no one accepts the question within ten minutes, it will be sent to a further ten mentors with the next highest ranking and so on and so on.
• 1-2-1 Tuition – The app also allows for a student to request a thirty-minute tuition session with a mentor.
  Much like the process in answering questions, the student will upload what they need tuition with and this will be sent on to ten mentors. The mentor who accepts will then have ten minutes to prepare before the tuition session starts over the app.
  The app provides audio and lesson recording for both parties with white board functionality if needed but as part of its security first build won’t allow for video calls or sharing.
• Feedback Requests – The final functionality of the app allows students to upload longer pieces such as essays or coursework to request feedback from mentors who will respond with a set of notes on the piece to improve it.

As mentioned, the app also contains a five-star rating system (similar to the one used for Uber drivers) that students can use to provide feedback on their mentors.

Part of the build process was to add functionality so that when new questions come in the mentors with the highest star rating would be sent the questions first so as to maintain the highest levels of accuracy and professionalism within the platform. 

As part of our Continuous Improvement ethos, once the app was built, cloudThing continued to work with the Peerz Academy to add automation and intelligent functionality to the platform.

Once mentors had started to answer questions, Peerz were keen to utilise this data to automatically surface answers to previously asked questions.

Through a combination of Azure's pre-trained Cognitive Services and an in-house exploratory analysis process, we developed a system which scans all the questions which have been uploaded to the system, and extracts all the salient information as structured data, which is stored in a Cognitive Search index.

When a new question is asked, it’s then scanned by the same system and used to query the search index to see if we already have a high quality answer to that exact question.

The system has been shown to retrieve exact answers with a precision of over 90% and to be extremely robust to edge cases like blurred images, handwritten questions, and questions with similar wordings.