Featured Projects
Four polished Android applications demonstrating architecture, security, AI integration, and user experience design
Notes App
Elevator Pitch
A production-quality note-taking app built in Java that goes far beyond a simple CRUD demo. It handles the full identity lifecycle — email verification, secure auth, password reset — then layers on image attachments stored in Firebase Storage, scheduled reminders that fire even when the app is closed, and seamless offline read access. Every architectural decision was made to show real-world reliability: what happens when the network drops, when the app is killed, when the user forgets their password.
Key Features
Tech Stack
Screenshots
NotificationWorker that survives app death and device restarts — unlike AlarmManager, it respects Doze mode and battery optimization constraints. Firestore offline persistence is enabled with multi-tab synchronization so reads are instant from local cache, while writes queue and sync automatically the moment connectivity returns. The full auth lifecycle (register → email verify → login → password reset) is handled without any third-party auth UI library.
Future Improvements
- Full-text search, tags & filters
- Export/import notes as JSON backup
- End-to-end encryption for sensitive notes
XChat
Elevator Pitch
XChat is a full-stack, one-to-one messaging app with three layers of real-time technology working simultaneously — Firebase for presence and message sync, Socket.IO for WebRTC signaling, and WebRTC itself for peer audio/video calls. All messages are encrypted end-to-end with AES keys derived from user credentials, so even Firebase cannot read the content. A built-in Gemini AI assistant lets users switch from human chat to AI chat in the same interface, with API keys protected from reverse engineering.
Key Features
Tech Stack
Screenshots
Future Improvements
- Group chat & group calls
- Signal-protocol-like double-ratchet key management
- TURN server deployment for better NAT traversal
HelpHive
Elevator Pitch
HelpHive is a three-app ecosystem designed to combat modern social isolation. HelpHive Core matches neighbors for practical assistance. MoodMosaic lets users log emotions, visualize aggregated community mood patterns on a neighborhood heat map, and receive clinically-informed AI support messages when negative emotions are detected. KindConnect captures everyday acts of kindness and surfaces them in a positive-only community feed — with no leaderboards, only intrinsic recognition. tested with 12+ users. It has offline sync reliability, built on clean architecture with Hilt DI and a layered Room/Firestore repository.
Key Features
Tech Stack
Screenshots
Future Improvements
- Verified volunteers & human moderation tools
- Anonymous mode for sensitive posts
- Integration with municipal emergency response systems
WebNovel Summarizer
Elevator Pitch
A specialized Android app that embeds a full-featured web browser — pre-configured for 6+ major novel platforms with a 15+ domain ad blocker and dark-mode CSS injection — and can summarize any chapter with a single tap. Three summary depths (5-second scan, 15-seconds standard, 25-seconds deep dive) are generated by Cerebras GPT-OSS-120B which is very fast and has excellent quality. For fall back, have Groq's Llama 3.3 70B integrated with custom system prompts that preserve honorifics, cultivation terminology, skill names, and character names exactly as written. An offline library organizes saved summaries into a Novels → Volumes → Chapters hierarchy for permanent access — no account, no cloud sync, zero data collection. Text-to-Speech with intelligent sentence-boundary chunking lets users listen to any summary hands-free.
Key Features
Tech Stack
Screenshots
ContentHolder singleton bypasses Android's Intent 1MB transaction limit when passing large chapter text to the summary screen. The TTS engine splits summaries at sentence boundaries into 4,000-character chunks with sequential playback and full pause/resume state preserved across chunks.
Future Improvements
- Batch summarization across multiple chapters at once
- Export library to PDF / EPUB formats
- Multi-model AI support (Claude, Gemini, GPT-4)