Students Channel Brain

An AI-powered study assistant that ingests lecture content, builds a RAG (Retrieval-Augmented Generation) pipeline, and serves an agentic tutor through a web app — with quizzes, a Q&A tutor, and weak-topic tracking. The project is benchmarked and evaluated end-to-end using LangSmith.

✨ Features

• Course Builder — ingest and structure lecture material into study modules
• Tutor Tab — conversational Q&A agent grounded in course content via RAG
• Quiz Tab — auto-generated quizzes based on lecture content
• Weak Topics Tab — tracks student performance to surface topics needing review
• Benchmarking & Evaluation — automated agent evaluation pipeline using LangSmith, scored across 38 tasks

🏗️ Architecture

See 05 - presentation/assets/Architecture and Data Flow.png for the full system diagram covering ingestion → RAG pipeline → agent → web app.

📁 Repository Structure

├── 00 - notebooks/
│   ├── 01_ingestion.ipynb            # Load & preprocess lecture content
│   ├── 02_rag_pipeline.ipynb         # Build the retrieval-augmented generation pipeline
│   ├── 03_agent.ipynb                # Agent logic (tutor, quiz, weak-topics)
│   ├── 04_web_app.ipynb              # Web app implementation
│   ├── 05_benchmark_langsmith.ipynb  # LangSmith benchmarking setup & runs
│   └── 06_final_evaluation.ipynb     # Final evaluation across the task set
│
├── 01 - benchmark/
│   ├── Benchmark_LangSmith_Logs.png
│   └── langsmith_benchmark_question_runs.csv
│
├── 02 - evaluation/
│   ├── Evaluation_LangSmith_Logs.png
│   ├── Lecture_Tutor_38_Task_Evaluation_Set.xlsx
│   └── Lecture_Tutor_38_Task_Final_Results.xlsx
│
├── 03 - reports/
│   ├── Benchmark_LangSmith_Analysis_Report.pdf
│   ├── Final_Evaluation_Report_38_Tasks.pdf
│   └── Students_Channel_Brain_Final_Project_Report.pdf
│
├── 04 - webapp/
│   ├── Course Builder 1.png
│   ├── Course Builder 2.png
│   ├── Quiz Tab.png
│   ├── Tutor Tab.png
│   └── Weak Topics Tab.png
│
├── 05 - presentation/
│   ├── assets/
│   │   └── Architecture and Data Flow.png
│   ├── sample-output/
│   │   └── Lecture 1: The Geometry of Linear Equations.html
│   └── The_Students_Channel_Brain_Presentation.pptx
│
├── requirements.txt
├── LICENSE
└── README.md

🚀 Getting Started

Prerequisites

Install dependencies listed in requirements.txt:

pip install -r requirements.txt

Usage

1. Ingest lecture content — run 00 - notebooks/01_ingestion.ipynb
2. Build the RAG pipeline — run 00 - notebooks/02_rag_pipeline.ipynb
3. Set up the agent — run 00 - notebooks/03_agent.ipynb
4. Launch the web app — run 00 - notebooks/04_web_app.ipynb
5. Benchmark with LangSmith — run 00 - notebooks/05_benchmark_langsmith.ipynb
6. Run the final evaluation — run 00 - notebooks/06_final_evaluation.ipynb

📊 Benchmarking & Evaluation

The agent was evaluated on a 38-task test set using LangSmith, with results and logs tracked in 01 - benchmark/ and 02 - evaluation/. Full write-ups are available in 03 - reports/:

• Benchmark_LangSmith_Analysis_Report.pdf — analysis of benchmark runs
• Final_Evaluation_Report_38_Tasks.pdf — detailed evaluation results
• Students_Channel_Brain_Final_Project_Report.pdf — full project report

🖼️ Screenshots

Web app UI screenshots (Course Builder, Quiz, Tutor, and Weak Topics tabs) are available in 04 - webapp/.

📘 Sample Output

05 - presentation/sample-output/Lecture 1: The Geometry of Linear Equations.html is an example HTML study module generated by the Course Builder, included as a reference artifact.

🎤 Presentation

05 - presentation/The_Students_Channel_Brain_Presentation.pptx is the graduation presentation deck, covering the problem, architecture, agent design, benchmark results, scale-up decision, final evaluation, failure analysis, and next steps.

📄 License

This project is licensed under the MIT License.