Selected
Projects
Real products I've designed, built, and scaled
PocketMind
Offline-First Personal AI
A premium, offline-first personal AI assistant mobile application that runs high-performance LLMs entirely on-device. It allows users to query their local documents (PDF, DOCX, Excel) and images (via OCR) in a 100% private environment without sending data to external servers.
On-Device LLM Inference
Embedded fllama engine (llama.cpp port) executing quantized GGUF models on-device with optimized contexts up to 4096 tokens
Smart Document RAG
Parses PDF, DOCX, and Excel formats locally to inject text dynamically into the model's chat context
On-Device OCR Engine
Integrates Google ML Kit for offline image text transcription and processing directly on-device
Offline-Safe Sync Queue
Buffers analytics and engagement events locally in SharedPreferences, batch syncing to Firestore and Sheets when online
Flutter
Dart
C++
Android
FirebaseGoogle ML KitHeartEcho
AI Companion Platform
I built the entire backend from scratch — a custom Priority Queue for AI scheduling, Redis Emotional Cache with sub-12ms retrieval, and Nginx load balancing with sticky sessions across 4 Node.js clusters.
Custom Priority Queue
Per-user AI scheduling with fairness guarantees and backpressure control
Redis Emotional Cache
Hash-indexed context retrieval in <12ms, TTL-based expiry strategy
Nginx Sticky Sessions
Consistent routing across 4 Node.js instances, zero session loss
WebSocket Messaging
Socket.io real-time, sub-100ms at 550K concurrent users
Next.js
Node.js
OpenAI
Redis
Nginx
MongoDB
Socket.io
AWSPaperXify
AI Note Generator
Parallel LLM pipelines chunk transcripts and call OpenAI simultaneously. Puppeteer renders pixel-perfect PDFs with auto table of contents — entire pipeline completes in <8 seconds.
Parallel LLM Pipeline
Simultaneous OpenAI calls per transcript chunk for max throughput
Semantic Chunking
Natural boundary detection for topic-aware note segmentation
Puppeteer PDF Render
Pixel-perfect formatted output with auto table of contents
Google Drive API
One-click save directly to user's cloud storage folder
MERNNode.jsOpenAI
PuppeteerMongoDB
TypeScriptSocial Manager
Autonomous Content Engine
Monitors real-time trends on Instagram and X, generates brand-aligned content via AI agents, and orchestrates publishing across platforms — entirely without human intervention.
Real-time Trend Analysis
Continuous monitoring of Instagram & X for emerging viral patterns
AI Agent Orchestration
Multi-step reasoning pipeline for brand-voice-aligned content
Kafka Event Bus
Decoupled producer/consumer for reliable, ordered scheduling
Multi-platform Publishing
Automated posting with platform-specific format and timing
Node.js
KafkaRedisTypeScript
DockerAWS
MicroservicesThis Portfolio
Next.js + MDX + Paper Theme
Next.js App Router + MDX for blog posts, case studies, and research. A fully custom sketchbook design system — paper textures, ink borders, handwriting fonts — with scroll-driven Framer Motion animations.
Next.js App Router
React Server Components for near-instant page loads
MDX Blog Engine
Markdown + JSX, compiled at build time — zero CMS, zero database
Custom Design System
Paper texture, ink borders, Architects Daughter handwriting font
Live GitHub Heatmap
Real contribution data from public API, refreshes daily
Next.jsTypeScript
MDX
Framer Motion
Tailwind CSSPocketMind
Offline-First Personal AI
A premium, offline-first personal AI assistant mobile application that runs high-performance LLMs entirely on-device. It allows users to query their local documents (PDF, DOCX, Excel) and images (via OCR) in a 100% private environment without sending data to external servers.
On-Device LLM Inference
Embedded fllama engine (llama.cpp port) executing quantized GGUF models on-device with optimized contexts up to 4096 tokens
Smart Document RAG
Parses PDF, DOCX, and Excel formats locally to inject text dynamically into the model's chat context
On-Device OCR Engine
Integrates Google ML Kit for offline image text transcription and processing directly on-device
Offline-Safe Sync Queue
Buffers analytics and engagement events locally in SharedPreferences, batch syncing to Firestore and Sheets when online
FlutterDartC++AndroidFirebaseGoogle ML KitHeartEcho
AI Companion Platform
I built the entire backend from scratch — a custom Priority Queue for AI scheduling, Redis Emotional Cache with sub-12ms retrieval, and Nginx load balancing with sticky sessions across 4 Node.js clusters.
Custom Priority Queue
Per-user AI scheduling with fairness guarantees and backpressure control
Redis Emotional Cache
Hash-indexed context retrieval in <12ms, TTL-based expiry strategy
Nginx Sticky Sessions
Consistent routing across 4 Node.js instances, zero session loss
WebSocket Messaging
Socket.io real-time, sub-100ms at 550K concurrent users
Next.jsNode.jsOpenAIRedisNginxMongoDBSocket.ioAWSPaperXify
AI Note Generator
Parallel LLM pipelines chunk transcripts and call OpenAI simultaneously. Puppeteer renders pixel-perfect PDFs with auto table of contents — entire pipeline completes in <8 seconds.
Parallel LLM Pipeline
Simultaneous OpenAI calls per transcript chunk for max throughput
Semantic Chunking
Natural boundary detection for topic-aware note segmentation
Puppeteer PDF Render
Pixel-perfect formatted output with auto table of contents
Google Drive API
One-click save directly to user's cloud storage folder
MERNNode.jsOpenAIPuppeteerMongoDBTypeScriptZero human input — trend analysis, content generation, and publishing on autopilot
Social Manager
Autonomous Content Engine
Monitors real-time trends on Instagram and X, generates brand-aligned content via AI agents, and orchestrates publishing across platforms — entirely without human intervention.
Real-time Trend Analysis
Continuous monitoring of Instagram & X for emerging viral patterns
AI Agent Orchestration
Multi-step reasoning pipeline for brand-voice-aligned content
Kafka Event Bus
Decoupled producer/consumer for reliable, ordered scheduling
Multi-platform Publishing
Automated posting with platform-specific format and timing
Node.jsKafkaRedisTypeScriptDockerAWSMicroservicesThis Portfolio
Next.js + MDX + Paper Theme
Next.js App Router + MDX for blog posts, case studies, and research. A fully custom sketchbook design system — paper textures, ink borders, handwriting fonts — with scroll-driven Framer Motion animations.
Next.js App Router
React Server Components for near-instant page loads
MDX Blog Engine
Markdown + JSX, compiled at build time — zero CMS, zero database
Custom Design System
Paper texture, ink borders, Architects Daughter handwriting font
Live GitHub Heatmap
Real contribution data from public API, refreshes daily
Next.jsTypeScriptMDXFramer MotionTailwind CSSAll open-source repos on GitHub