Eleceed

Eleceed – The AI Operating System for Student Success

Comment

Eleceed – AI-Powered Learning, Assessment & Placement Platform

🚀 Inspiration

Students preparing for placements, JEE, GATE, TNPSC, UPSC, and other competitive examinations often use multiple disconnected platforms for learning, assessments, coding practice, and mock interviews. Institutions face a similar challenge—creating high-quality assessments, tracking student progress, and providing personalized guidance at scale.

We built Eleceed to solve this problem by creating a unified AI-powered platform that combines question generation, assessments, AI interviews, and knowledge-based learning into a single ecosystem. Our vision is to provide every student with an intelligent learning companion while enabling institutions to automate training and placement preparation.

🛠️ How We Built It

Eleceed is built using a modern AI-first architecture:

Frontend

  • Next.js
  • React
  • Tailwind CSS
  • TypeScript

Backend

  • NestJS
  • PostgreSQL
  • pgvector
  • REST APIs

AI Stack

  • Gemini / OpenAI compatible LLMs
  • Retrieval-Augmented Generation (RAG)
  • LiveKit Realtime Voice Agents
  • Vector Search using pgvector

Platform Workflow

Student / Faculty
        │
        ▼
 Next.js Dashboard
        │
        ▼
    NestJS APIs
        │
 ┌──────┴──────┐
 ▼             ▼
PostgreSQL   AI Services
+ pgvector   + Voice Agents

Faculty members can upload PDFs, notes, question banks, websites, and study materials. These resources are converted into vector embeddings and stored inside PostgreSQL using pgvector.

When a student requests a quiz, mock interview, or assessment, Eleceed retrieves the most relevant learning content and generates personalized questions.

The relevance score is calculated using cosine similarity:

$$ Similarity(q,d)=\frac{q \cdot d}{||q|| \times ||d||} $$

where:

  • (q) = query embedding
  • (d) = document embedding

This allows the platform to retrieve contextually relevant information before generating questions and interview scenarios.

✨ Features

AI Question Generator

Generate:

  • MCQs
  • Descriptive Questions
  • Coding Challenges
  • Competitive Exam Questions

AI Mock Interviews

Students can participate in real-time voice interviews that evaluate:

  • Technical knowledge
  • Communication skills
  • Problem-solving ability
  • Behavioral responses

Assessment Engine

  • Timed tests
  • Automatic evaluation
  • Performance analytics
  • Faculty review workflows

Knowledge Base

Create searchable knowledge repositories from:

  • PDFs
  • Documents
  • Websites
  • Study materials

Placement Readiness Tracking

Track student performance across:

  • Assessments
  • Mock interviews
  • Coding exercises
  • Learning activities

📚 What We Learned

RAG Is More Important Than Prompting

Initially, we focused heavily on prompts. We later discovered that retrieval quality has a much larger impact on output quality than prompt engineering alone.

AI Systems Need Strong Architecture

Separating AI services from frontend applications allowed us to switch between different models without affecting the user experience.

Real-Time Voice Is Challenging

Building conversational AI interviews required handling speech recognition, audio streaming, latency optimization, and browser compatibility.

Multi-Tenant Design Matters

Educational institutions have different departments, courses, and learning structures. Designing a scalable hierarchy from the beginning helped avoid major architectural changes later.

🧗 Challenges We Faced

Challenge 1: Low-Quality Question Generation

Early versions generated simple factual questions that lacked depth.

Solution: We introduced curriculum-aware RAG pipelines and Bloom's Taxonomy-based prompts to generate more analytical and examination-focused questions.

Challenge 2: Retrieval Performance

As the number of uploaded documents increased, retrieval latency grew significantly.

Solution: We implemented pgvector indexing and optimized similarity search to retrieve relevant content efficiently.

Challenge 3: Real-Time Voice Conversations

Speech recognition and AI responses occasionally caused delays and interruptions.

Solution: We built event-driven voice workflows using LiveKit and optimized state synchronization between the browser and backend services.

Challenge 4: Assessment Consistency

Faculty members often updated assessments while students were actively taking tests.

Solution: We implemented assessment versioning to preserve historical attempts while allowing future edits.

🎯 Impact

Eleceed transforms learning from a passive experience into an interactive and personalized journey. By combining AI-generated assessments, real-time mock interviews, and knowledge-aware learning, the platform helps students prepare more effectively for placements and competitive examinations.

Our long-term vision is to become the AI-powered operating system for education and career readiness, enabling institutions to train thousands of students efficiently while delivering personalized learning experiences at scale.

Built With

Share this project:

Updates

Submission history