In the ever-evolving landscape of education technology, our client, a startup backed by funding from the Gates Foundation, embarked on a transformative journey. Our collaboration aimed at enhancing student’s learning experience through an innovative e-learning platform.

The existing application is a platform where students can write essays and receive feedback and guidance on writing structure. At the start, the application could only provide feedback and a grade. Our first stepping stone was to enhance an inefficient ML solution for grammar correction. Later, we worked on revamping the UI and implemented the guidance functionality, using more ML models.

The original code underwent a thorough rewrite that showcased the collective talents and efforts of our client’s seasoned developers and our dedicated team at HEITS. Our biggest milestone was to rebuild the app from the ground up and rewrite it using Python and JavaScript. All this was done while also maintaining the application functions for existing users.

This endeavor forged a collaborative spirit that elevated both teams. Following this significant milestone, the seamless integration of skills and expertise from both sides earned our HEITS team recognition as a "fire-and-forget" team, reflecting the mutual trust in the capabilities and contributions of both teams.

Adhering to agile methodologies like scrum facilitated seamless development, with a particular focus on crafting the app’s Chrome extension.

To align with the target audience's usage of Google Docs, we strived to provide a solution compatible with the client's preferences. A notable cost-saving initiative involved replacing a cache-based feature, which cost $2000/month, with a machine-learning solution, which resulted in a reduction of expenses to tens of dollars.

We tapped into the power of machine learning to recognize intricate word patterns, enabling us to classify sentences into distinct categories such as Thesis, Analysis, Summary, and more. This feature empowers educators and students with a nuanced understanding of student submissions.

Our system goes beyond mere classification; it facilitates comparison, allowing us to gauge whether a student remains on topic or steers off course. This ensures a more refined evaluation of submissions, emphasizing relevance and coherence.

The ML system looks over the text the student provides and analyzes it. The system looks at the words in a sentence and based on some key elements, shows a percentage for the types of sentences it matches. Based on percentage, we show the user the most accurate label for that set sentence or piece of text.

Teachers seeking a preview of assignment feasibility can leverage our platform's capability to generate essay examples. This feature not only aids educators in assessing the approachability of assignments but also fosters a proactive teaching environment.

To streamline the assignment creation process, we've incorporated a prompt suggestion feature. Teachers can benefit from our system's ability to generate relevant prompts, facilitating a smoother assignment development experience.

Delving deeper into linguistic intricacies our technology excels at extracting grammar details, word dependencies, and co-referencing information within and between sentences. This meticulous analysis contributes to a comprehensive understanding of language structures, enhancing the overall quality of assessments.

Addressing challenges from competitors utilizing ChatGPT, a proof of concept emphasized explainability. This involved formalizing rules from student-written sentences to guide the writing process.

Constant optimization efforts ensure that, despite numerous ML models, execution time remains consistently under 5 seconds. Iterations and improvements are ongoing, reflecting a commitment to delivering a seamless user experience.

Although the product is just now starting to spread its wings and reach more schools, its impact on students has been substantial.

The platform's launch and ongoing integration into schools showcase its potential impact. Discussions with administrators, approval processes, and live demos contribute to a positive trajectory.

Teachers now have a dashboard, and the application supports text authentication for comparing submissions.

The collaborative relationship between our team and the client has been instrumental. We don't merely execute; we actively contribute to architectural decisions, estimates, and development. We communicate with the client and participate in decisions, advising on the best course of action when technology comes into place, but not only that, the team is eager to help improve our features constantly. It comes up with proposals on how to do that. We work in sprints, plan, and work together to achieve the project goals.

For the project to be successful we also use analytics to advise us on the course we are taking, helping us see what is working for our customers and where the value is.

This project represents a remarkable transformation from a scenario of scattered knowledge and outdated code. With a cohesive team, agile development processes, and strategic collaboration with the client, we've successfully navigated challenges, paving the way for a groundbreaking educational platform. As we continue to iterate and improve, our collaboration stands as a testament to the power of innovation and effective teamwork in reshaping the future of education.

Web Team

1. • React
2. • NodeJS
3. • ExpressJS
4. • Typescript
5. • Terraform
6. • Terragrunt
7. • MySQL
8. • AWS
9. • SQS
10. • Chrome Extensions

1. • OpenAI API
2. • Langchain
3. • SpaCy
4. • Custom LLMs (Distilled Bert, MPnet)
5. • PyTorch for training models
6. • ONNX for deployment models

1. • Redis
2. • AWS
3. • SQS
4. • MySQL
5. • Flask