AI Quiz & Summaries
Budget: ₹1,500 – ₹12,500 INR
The core of my learning platform is already live: a React front end talking to a Node-Express API with PostgreSQL, secured by JWT, and enriched with a custom Rich-Text Editor, ECharts dashboards, and IndexedDB offline mode. Where I now need your help is the AI layer that creates quizzes and content summaries through Groq and Google GenAI.
I want to push this module further in three directions at once—accuracy, new feature depth, and raw performance—without disrupting the existing users.
Scope of work
• Audit the current prompt engineering, model selection, and caching strategy, then tighten it so the generated questions better match source material and difficulty settings.
• Extend functionality: additional question types (e.g., drag-and-drop, image-based), adjustable Bloom-taxonomy levels, and richer summary formats with highlights and key-takeaway flash cards.
• Streamline response times and cost: optimize batching, parallel requests, and intelligent result caching; profile Node/Express endpoints and fine-tune PostgreSQL queries to remove bottlenecks.
Acceptance criteria
• At least a 20 % decrease in average generation latency under load testing.
• Observable accuracy lift demonstrated through an evaluation set I provide.
• All new endpoints fully covered by unit and integration tests and merged via pull request to the existing GitHub repo with CI passing.
You will be working in a TypeScript, React 18, Node 20, PostgreSQL 15 stack, and I can spin up a staging database plus OpenAI/Groq keys for isolated testing. If you have experience fine-tuning or function-calling these models, even better—let’s make this AI engine shine.
I want to push this module further in three directions at once—accuracy, new feature depth, and raw performance—without disrupting the existing users.
Scope of work
• Audit the current prompt engineering, model selection, and caching strategy, then tighten it so the generated questions better match source material and difficulty settings.
• Extend functionality: additional question types (e.g., drag-and-drop, image-based), adjustable Bloom-taxonomy levels, and richer summary formats with highlights and key-takeaway flash cards.
• Streamline response times and cost: optimize batching, parallel requests, and intelligent result caching; profile Node/Express endpoints and fine-tune PostgreSQL queries to remove bottlenecks.
Acceptance criteria
• At least a 20 % decrease in average generation latency under load testing.
• Observable accuracy lift demonstrated through an evaluation set I provide.
• All new endpoints fully covered by unit and integration tests and merged via pull request to the existing GitHub repo with CI passing.
You will be working in a TypeScript, React 18, Node 20, PostgreSQL 15 stack, and I can spin up a staging database plus OpenAI/Groq keys for isolated testing. If you have experience fine-tuning or function-calling these models, even better—let’s make this AI engine shine.