Inspiration
We were inspired by the challenges faced by teachers in environments with high student-teacher ratios, particularly in government and budget private schools. Teachers are overburdened with grading and providing individual feedback, leaving little time for mentorship. We wanted to create a solution that empowers educators while ensuring students receive the personalized attention they deserve. Our goal aligns with UN Sustainable Development Goal 4: ensuring inclusive and equitable quality education and promoting lifelong learning opportunities for all.
What it does
Lumen Slate uses both its Solo Agents and Independent Agents in collaboration on the platform, to help teachers reduce their time taken in non teaching activities like - quiz creation, grading/assessment, personalized content generation and more.
How we built it
We built Lumen Slate using a modular web-based architecture that leverages cutting-edge GenAI tools like Gemini and Vertex AI.
🧠 Overall System Architecture
Our architecture combines frontend interfaces, AI microservices, and cloud infrastructure to deliver a seamless educational experience.
🤖 Solo Agents Architecture
Solo agents handle specific, independent tasks such as question generation or answer evaluation using dedicated AI pipelines.
🕸️ Orchestrated Agents Architecture
In orchestrated mode, multiple agents collaborate to complete complex workflows like voice-based assessment creation and personalized feedback generation.
Challenges we ran into
We faced challenges regarding deployment of the gin backend, the gRPC microservice and the finetuning of the responses of multiple ADK-based AI Agents. It was tricky to handle their structured outputs and then use them to communicate across microservices for database and miscellaneous operations.
Accomplishments that we're proud of
We are proud of creating a solution that meaningfully reduces teacher workload while enhancing student learning experiences. Our AI-driven personalization, ability to generate unique questions, and progress reporting are standout features that differentiate Lumen Slate from existing platforms like Google Classroom or Moodle.
What we learned
This was our first tiem using bolt, its is far from making production ready apps yet, but we enjoyed its simplicity and how it managed to weed out the labour intensive parts fo the project. We have experience with using Langchain and LangGraph but this was our first time with ADK and its simplicity really impressed us. We gained experience in building AI microservices, integrating multimodal inputs, optimizing an agentic workflow in ADK.
What's next for Lumen Slate
We plan to make our agents better, less prone to hallucination - and better pipelines. Since we have written our code in Google's Flutter framework, we plan to make our app work for android and IOS platforms as well.
Log in or sign up for Devpost to join the conversation.