Most students don't have a note-taking problem. They have a fragmentation problem. By the middle of a semester, course material ends up scattered across lecture slides, PDFs, recorded lectures, screenshots, YouTube videos, ChatGPT conversations, Google Docs, Notion pages, and random browser tabs you'll definitely come back to later.
When it comes time to study, finding information becomes harder than understanding it. The modern student has effectively become their own search engine. SaveMyGrade started from a simple observation: what if every piece of academic information entered the same system? Not your notes in one place. Not your PDFs in another. Not your AI conversations somewhere else. Everything. The goal wasn't to build another study application. The goal was to build a living academic memory.
The product philosophy
The workflow for most students today: download lecture slides, upload them to ChatGPT, ask questions, close the conversation, repeat next week. The context disappears. The model forgets. You start over. Every single time.
SaveMyGrade takes the opposite approach. Every PDF, video, image, lecture recording, and note becomes part of a continuously growing knowledge base. By week twelve, the application understands twelve weeks of material. By finals week, your entire semester exists inside a searchable representation of your coursework. When you sit down to study, you open one application — not Canvas, not ChatGPT, not Google Drive, not Notion.
Retrieval ends up being the product
Most AI applications today optimize generation. Better prompts. Better models. Better outputs. I found that retrieval mattered far more. The quality of an answer is largely determined before the language model generates a single token. Finding the right context became the core engineering problem: ingestion → chunking → embeddings → vector search → reranking → generation → citations.
A good answer isn't produced by a smarter model. A good answer is produced by giving the model the right context.
A search engine for your own brain
At its core, SaveMyGrade is less of an AI chatbot and more of a personal search engine. Every piece of information goes through an ingestion pipeline — content extraction, chunking, embedding generation, metadata enrichment, indexing — producing a vector representation of your semester. Instead of searching filenames or folders, you search concepts. “Where was that slide about TCP congestion control?” becomes “Explain TCP congestion control and show me where I learned it.” That distinction ends up being surprisingly powerful.
Multimodal means more than images
Most applications advertise multimodal support. Often this means an image receives an embedding vector and can be retrieved later. That isn't enough. A photo of a whiteboard diagram should contribute to chat answers, flashcards, summaries, concept relationships, and knowledge graphs. The system performs additional analysis during ingestion so that images become understandable context rather than binary files sitting in storage. The objective is not storing media. It's extracting knowledge from media.
Local-first AI is underrated
Most AI applications immediately push user data into the cloud. SaveMyGrade runs entirely locally. Setup: clone the repository, add a Gemini API key, start the application. That's it. No accounts. No subscriptions. No external databases. No cloud infrastructure. There is something refreshing about software that simply starts working after configuration instead of onboarding you through five SaaS dashboards.
Building with AI changes what matters
This project also became an experiment in modern software engineering. Implementation is becoming cheaper. Architecture isn't. Writing an endpoint is easy. Changing the ownership model of your data isn't. The limiting factor increasingly feels less like code generation and more like maintaining conceptual integrity across an evolving system.
Looking forward
The most exciting part about personal AI systems isn't generation. It's memory. Language models are becoming increasingly capable at reasoning. What they still lack is context. Your context. Your notes. Your lectures. Your understanding of the world. SaveMyGrade is an attempt to bridge that gap. Not another chatbot. Not another notes app. A living brain for everything you learn throughout a semester — and perhaps eventually, everything you learn beyond it.
View SaveMyGrade on GitHub
Vansh Sinha is a Tempe-based
Project Index
Spotify
