Tech: Next.js 14 TypeScript Convex OpenAI TTS DALL-E 3 Clerk Tailwind CSS Shadcn UI

• Engineered AI-driven SaaS enabling text-to-podcast generation with 6 voice models via OpenAI TTS API
• Integrated DALL-E 3 for dynamic thumbnail generation from natural language prompts
• Implemented Clerk-based authentication with secure session management and protected API routes
• Built a fixed-position audio player with seek, volume, mute — persists across page navigation
• Designed a real-time discovery engine with search, filtering by voice type, and trending surfacing
• Deployed on Convex for real-time backend with automatic data sync and serverless functions

Tech: Python FastAPI Streamlit LangChain LangGraph

• Built a multi-turn AI conversational system using LangChain with structured evaluation of response quality
• Architected a state-graph workflow with LangGraph for managing conversation context and therapeutic patterns
• Developed real-time data processing pipelines to monitor system performance, response latency, and session metrics
• Implemented response quality metrics for reliability tracking and workflow efficiency benchmarking

Tech: Python arXiv API PyPDF2 Streamlit

• Designed an autonomous AI agent that browses, reads, and synthesizes papers into publication-ready output
• Implemented keyword-based ranking and semantic filtering achieving ~90% retrieval accuracy
• Built PDF extraction pipelines using PyPDF2 for processing large volumes of research documents at scale
• Automated end-to-end summarization workflows, reducing manual research effort by ~80%

Tech: React Groq (Llama 3.3 70B) Web Speech API Tailwind CSS Framer Motion

• Built a real-time voice interaction system using Web Speech API for ultra-low latency speech recognition
• Integrated Groq LPU inference (Llama 3.3 70B) for sub-second AI responses faster than human reactivity
• Implemented pedagogical scaffolding — Hook → Contextualization → Prompt → Gentle Correction
• Designed a lexile-appropriate AI persona with 5-7 word sentence constraints for age 3-5 interaction

Tech: React 19 Vite @dnd-kit Context API Vitest

• Engineered a 3-column Kanban board with full CRUD, drag-and-drop (via @dnd-kit), and localStorage persistence
• Built a real-time search dropdown with matching results, color-coded status dots, and priority badges
• Implemented centralized state management via Context API with memoized derived state
• Added an Error Boundary with automatic recovery UI and storage cleanup
• 12 passing tests using Vitest + React Testing Library covering auth, CRUD, and rendering

Tech: Django REST Framework React PostgreSQL Docker

• Solved the N+1 query problem for nested comments using O(1) in-memory adjacency-list tree reconstruction
• Implemented a 24-hour rolling karma leaderboard using split-and-merge aggregate queries
• Built polymorphic voting via GenericForeignKey with idempotent API and unique_together constraints
• Designed a recursive React UI handling unlimited comment nesting without performance degradation

Tech: Node.js REST API Batch Processing

• Built a Node.js client aggregating data from 500 IoT devices against a rate-limited API (1 req/sec)
• Implemented intelligent batching (groups of 10) reducing total API calls from 500 → 50
• Designed exponential backoff retry logic for automatic recovery from 429 responses
• Serial processing with deliberate delays to respect API limits while maintaining data integrity