Hi! I created SolariStudy, basically a specialised study platform and productivity app (I built it with React Native, Expo and Supabase). The app addreses a problem which I face a lot every day and a lot of students preparing for rigorous exams and olympiads face it too: standard AI tools tipically just generate big quizzes in a MCQ format or just give immediate solutions that kind of ruin the learning process (also cluttering interfaces with unorganized chat logs). My idea for this app was to try a dual-mode engine which features SolariLearn for normal tutoring (just perfect for going through study materials or books) and SolariSolve for strict, timed problem sessions (very useful for Olympiad preparation especially).
Here I introduced major QoL improvements that structure the revision workflow, including organizing sessions into dedicated, isolated subject folders (like Mathematics, Physics, Linguistics, Astronomy etc.) to prevent the cluttering of chats (one of the problems mentioned earlier, which is solved by a few AI’s, but the process of organizing is not automated). The app also implements automated worksheet ingestion to extract and prepare textbook problems, introducing exam focus sprint timers to enforce accountability. By using the SolariSolve mode, you are able to get step-by-step scorecards with streak and accuracy tracking.
The most challenging tech hurdle was engineering the streaming AI pipeline (faced a lot of bugs here and did not really know where to start from), but also mathematical rendering ragebaited me at the end… Resolving API dropouts and rate limits required me to implement an exponential backoff with a dynamic model fallback system, alongside the real-time latex rendering to display the math formulas (which were essential for physics, math and astronomy so I couldnt just use normal text).
I am proudest of turning this small concept into a functional , working cross-platform project that I feel balances an unstimulating focus environment with rigorous grading (which is objectively done by the AI tool).
Now, for testing, you should know that while a compiled Android APK is abvailable for mobile testing, running the web version locally on pc is kinda my own recommendation :) since the mobile part was more of an experimental preview which can still receive UI polishing in the future. Most of the instructions are in the READme, but I can paste them here again. Local setup required node.js, cloning the repository, running npm install, supplying supabase and gemini api keys in an .env file and launching the project by entering in terminal npx expo start -w to test the engine directly in a web browser.
I really hope you like this project and that you can maybe use it for your own personal use.
- 5 devlogs
- 18h
- Frictionless

