Mobile + Backend
MindMitra
Cognitive-wellness app (iOS + Android)
MindMitra (formerly CognizenX) is a cognitive-wellness app built with DePauw Neuroscience. I led a three-engineer team decoupling AI and data workflows from the shipped React Native client onto a dedicated Vercel/MongoDB backend—isolating ML updates from mobile release cycles while delivering iOS and Android builds with telemetry-driven spaced-repetition scheduling.
- Role
- Software Engineer – MindMitra
- Team size
- 3 engineers
- Timeframe
- Aug 2025 – May 2026
- Status
- Shipped v2 (iOS + Android)
Previously released as CognizenX on the App Store; the product is now branded MindMitra.
Product screenshots
Context and problem
A shipped React Native cognitive-wellness app needed a backend-first architecture, Android support, server-bound AI generation, and quiz telemetry for research-informed spaced repetition.
Challenge
Take over a live v1 product under its prior CognizenX branding, decouple AI from the mobile client, expand to Android, and instrument telemetry for spaced-repetition research workflows without disrupting existing users.
Constraints
- Existing users and shipped iOS codebase to refactor without breaking core quiz flows
- Early deployment scope—not a large-scale production rollout
- Cross-platform release differences between iOS (TestFlight) and Android (Play internal track)
- App Store listing still published under the prior CognizenX name during the MindMitra rebrand
Key contributions
- Led a 3-engineer team to migrate AI and data workflows from a shipped React Native app into a dedicated backend architecture, enabling independent backend releases and faster iteration on ML-powered features.
- Built and maintained a 25-endpoint Node.js/Express backend on Vercel and MongoDB Atlas to centralize authentication, AI content generation, and scheduling workflows behind a unified API layer.
- Developed an automated GPT-4 content pipeline that generated 3,000+ questions across 40+ subdomains and ran ~300,000 embedding-based similarity checks to enforce semantic deduplication.
- Improved backend security and maintainability by moving exposed AI credentials server-side, implementing rate limiting and request validation, and expanding test coverage to 65 automated tests (~60% coverage).
Architecture
- Step 1React Native clientCross-platform mobile app (iOS + Android)
- Step 2Node.js / Express API25 endpoints on Vercel—auth, content generation, learning state, telemetry
- Step 3GPT-4 content pipeline3,000+ questions across 40+ subdomains with embedding deduplication
- Step 4Spaced-repetition telemetryQuiz performance data driving scheduling algorithms
Key engineering decisions
- Migrated AI and data workflows onto a dedicated backend to enable independent releases and faster ML iteration
- Moved exposed AI credentials server-side with rate limiting and request validation on backend routes
- Invested in embedding-based deduplication and automated test coverage (~60% backend) before scaling content generation
Security and reliability
- Embedding-based semantic deduplication across ~300,000 similarity checks to reduce redundant GPT-4 calls
- Rate limiting, request validation, and session authentication on backend routes
- 65 automated tests achieving ~60% backend code coverage
Testing
- 65 automated tests with ~60% backend code coverage
Outcome
Delivered MindMitra v2 on iOS and Android with a decoupled backend, 25-endpoint API, GPT-4 content pipeline (3,000+ questions), and telemetry-driven spaced-repetition scheduling.
Limitations
- Early deployment—not a large-scale production rollout
- iOS App Store listing remains under the prior CognizenX URL during rebrand
Next steps
- Complete MindMitra rebrand across app store listings
- Expand per-question progress tracking for spaced-repetition analytics
Tech stack
- React Native
- Node.js
- Express
- MongoDB Atlas
- GPT-4
- Vercel
- Jest
- Supertest