Mentor AI

Our solution is a AI speech assessment platform that allows for self-assessment of spoken presentations in the English language. Our solution is currently active in university settings for international students and we propose to adapt our solution for K-8 students, especially those who are immigrants or refugees. The assessment gives qualitative feedback regarding the learner's pronunciation, grammar and vocabulary among other metrics. It is used by English departments to offer students new ways of practicing speaking at home and to be guided without the teacher's presence.

Our solution provides ways for personalized learning path for each student and helps individuals improve on specific aspects of their speaking skills without the immediate need for a teacher.

This will mean that students can accelerate their path to full English fluency, leading to greater learning opportunities in other classes and improved opportunities to develop rapport with peers.

Increased access for disadvantaged students will mean greater opportunities to educational development that compounds with earlier English acquisition. The ultimate impact is enhanced engagement and motivation for school-life and education for immigrant and refugee students. 

The solution is already in place in university setting for international students. In terms of product delivery and technical capability, we are more than capable. For the communities, we are working with immigrant adult communities with more than 1000 members from the USA in our online free English classes. Although our product has not been adapted for the use of K-8 students, many of our immigrant community members are parents and have children who are learning English. The team lead is a former middle-school English teacher in Tunisia and concern for disadvantaged students one of the team leader's core values. We are ready to work with the Challenge partners to build and deliver a solution that makes the most sense for the communities they serve. 

We are completing our 4 month pilot and launching in May this year. 
We serve more than 4000 learners and 100 educators. 
With the Challenge, we will adapt our product for the K-8 audience by engaging with experts and Challenge partners. 

We have three aspects to our solution. Firstly, it allows for private practice for the learner and private feedback as well. It means the learner can be at ease when practicing speaking. At the same time, the assessment can be shared with a teacher who can then provide further granular feedback. The highly customizable tasks allows the teacher to guide the students up-close when required. Finally, the accuracy of our platform has been tested by IELTS examiners, validating the assessment criteria and qualitative comments with quantitative aspect (we use the scale: VERY GOOD, SATISFACTORY and NEEDS IMPROVEMENT). This saves teachers time when conducting speech assessments. This also means cost-savings for the school. This accessibility ensures that underserved schools and communities can benefit from advanced educational tools that were previously out of reach.

We're using multiple AI technologies:

• NLP is employed to analyze students' written and spoken responses. This allows for real-time feedback on language use, grammar, pronunciation, and comprehension.
• ML: Our models are primarily developed in-house using frameworks like TensorFlow and PyTorch, supplemented by pre-trained models where applicable to enhance performance and accuracy.
• LLM: We query LLM for some aspects of our report generation. We use third-party LLM solutions but are capable of deploying our own solution if needed.
• Speech to text: We integrate third-party speech recognition services that are robust and widely validated for accuracy and speed, ensuring a seamless user experience.
• Web-app: ReactJS front-end, Python back-end, deployed on all three major cloud service providers.

Tthe accuracy of our platform has been tested by IELTS examiners, validating the assessment criteria and qualitative comments with quantitative aspect (we use the scale: VERY GOOD, SATISFACTORY and NEEDS IMPROVEMENT). The value of the platform in improving learner experience has been attested to by students rating the platform 4.5 out of 5. 

You can see some feedback from students here:

To ensure equity and combat bias in our AI implementation, we focus on a comprehensive strategy that includes:

• Diverse Data Collection: We gather and analyze data from diverse demographic groups, ensuring our AI models are trained on datasets that reflect the backgrounds and experiences of all learners.
• Stakeholder Engagement: We actively engage with educators, parents, and community leaders from marginalized groups to tailor our solutions to the specific needs and challenges of these communities
• Bias Detection Mechanisms: Our development includes routine checks for bias in AI decision-making processes, using techniques like re-sampling to ensure fair representation in training data.
• Transparency and Accountability: We maintain transparency in our AI algorithms, allowing stakeholders to understand how decisions are made and facilitating trust and accountability.

We have 4 full-time members and 2 part-time.

2 full-time software engineers 

1 part-time full-stack developer

1 full-time Product manager

1 part-time UX UI product designer

1 full-time business development officer