projects / 2025 / Academic project
StudyForge — Agentic Document Intelligence Platform
A two-service document platform with RAG chat, YouTube-to-notes and text-to-slides, routing inference across Groq and self-hosted NVIDIA NIM.
What it is
A two-service platform built without an agent framework (the orchestration is hand-rolled). On top of one normalised document pipeline for DOCX, PPTX and XLSX files, it offers:
- RAG chat over your documents, backed by Supabase with pgvector;
- YouTube → notes;
- Text → PPT generation.
How it's built
- Inference routing: requests go to Groq or to a self-hosted NVIDIA NIM model behind a single abstraction layer, so the service code doesn't care which one answers.
- Semantic caching: similar questions reuse earlier answers; the cache hit rate is about 50%.
- Evaluation: RAGAS across 300+ automated test queries.
- Storage: files on AWS S3.