- 💬 Ask me about React, Angular, Typescript, Javascript, Java, Spring Boot, Node.js, Flask, FastAPI, Tkinter, Python and AI Agents.
- 📫 How to reach me karthikpotti842@gmail.com
- 👨💻 Know about me 🌐 karthikp-mac-portfolio
- 🚀 Google Developer Profile g.dev/karthikPMac
- 🤗 AI Projects (Live) huggingface.co/KarthikP-Mac/spaces
- 🐧 Automation & Scripting Linux automation using Java & WinSCP
- 📄 Resume (Live Preview) Karthik's Resume
- 📄 Resume (Markdown) Karthik's Resume
-
Web-RTC (Live Random Video Calls)
- Description: Real-time video calling application with media controls, peer-to-peer streaming, and screen sharing.
- Live Demo: web-rtc-lq00.onrender.com
- Tech Stack:
-
Web-Sockets (Live Chats)
- Description: Real-time instant messaging application ensuring reliable end-to-end delivery of chat messages.
- Live Demo: web-sockets-ju5x.onrender.com
- Tech Stack:
-
Sticky-Notes
- Description: Progressive Web App (PWA) designed as an interactive digital corkboard and kanban board, featuring free-form draggable note cards, coordinates persistence, custom tag categorizations, and multi-tab database sync.
- Live Demo: stickynotes-pk-mac.vercel.app
- Tech Stack:
-
BFSI-Insurance-Ai-Agent (Banking AI Copilot)
- Description: An intelligent finance & banking AI assistant that answers domain-specific queries. Leverages a robust CRAG pipeline with a custom recursive character splitter, FAISS vector database, and a Multi-Model LLM Orchestration architecture — configured with Qwen3.6 27B (qwen/qwen3.6-27b) as the active primary reasoning engine, with dynamic hot-swapping fallback routing across Llama-3.3-70b, Llama-3.1-8b, and GPT-OSS-120B to manage token limits and optimize cost per query.
- Live Demo: karthikp-mac-banking-ai-copilot.hf.space
- Tech Stack:
CRAGMulti-LLM - LLM Models Supported: Groq (Active: Qwen3.6 27B | Fallbacks: Llama-3.3-70b · Llama-3.1-8b · GPT-OSS-120B), Gemini, OpenAI, Claude.
-
Jarvis AI Assistant
- Description: A real-time, Multimodal Voice-to-Voice AI Assistant featuring dynamic location/timezone-aware greeting generation, dual-engine Text-to-Speech (TTS) synthesis, and high-performance bidirectional WebSocket communication.
- Frontend: Built with React.js (Vite), Vanilla CSS (Cyberpunk/Sci-Fi UI theme with active/inactive neon state cues), Web Audio API (AudioContext, AnalyserNode, GainNode for real-time frequency waveforms and chime synthesis), and the Web Speech API (webkitSpeechRecognition) for local background wake-word triggers.
- Backend: FastAPI ASGI server utilizing bidirectional WebSockets for high-frequency duplex streaming of raw audio packets and system configs.
- AI Models (STT & Multi-LLM): Powered by Groq API — Whisper-large-v3-turbo for rapid Speech-to-Text and a Multi-Model LLM Routing Layer selecting Qwen3.6 27B (qwen/qwen3.6-27b — active primary model) with fallback to openai/gpt-oss-120b (complex reasoning) and Llama-3.1-8b-instant (ultra-low latency) for agent reasoning, timezone-aware localized queries, and function-calling routing.
- Text-to-Speech (TTS): Dual-engine pipeline: Cloud/Premium (ElevenLabs API for voice streaming) and Local/Free (kokoro-onnx==0.3.4 and ONNX Runtime for in-memory offline English voice synthesis), with browser-native speech synthesis as a fallback for multi-lingual accents (Hindi & Telugu).
- Safety & Infrastructure: Double-layered regex guardrail filter running on the FastAPI server to block explicit or unsafe URLs and queries before LLM invocation. Deployed as a monolithic container on Hugging Face Spaces via Docker.
- Live Demo: karthikp-mac-jarvis-ai-assistant.hf.space
- Tech Stack:
- Description: A real-time, Multimodal Voice-to-Voice AI Assistant featuring dynamic location/timezone-aware greeting generation, dual-engine Text-to-Speech (TTS) synthesis, and high-performance bidirectional WebSocket communication.
-
Adarsh Computers Portal
- Description: A comprehensive educational portal built for an institution to manage courses, students, announcements and online quizzes.
- Live Website: adarsh-computers.vercel.app
- Frontend Tech:
- Backend Tech:
SMTP
-
First-Web-Site
-
Smart Stocker (Ledger Application)
- Description: Portable software for stock, ledger, and sales entry tracking built with Python's GUI framework.
- Tech Stack:
Application Output & Report Showcase:
Customer Bill 
Purchase Order 
Monthly Profit Report (Setup) 
Monthly Profit Report (Export) 
-
Python-Pdf Utility
- Description: Utility for generating PDFs from directories, selecting files/folders, and extracting text or images by selecting custom ranges.
- Tech Stack:
PyPDFReportLab
-
Auto Power Off Timer
- Description: Portable utility to easily schedule automatic sleep, shutdown, or restart commands on Windows PCs.
- Tech Stack:
-
Movie Seat Booking
-
DOM Array Methods Dashboard
-
Exchange Rate Calculator
-
Custom Video Player
-
Form Validator
-
Menu Slider & Modal
-
Calculator
-
JSON Card Generator
-
Click-to-Toggle Visibility UI
-
HTML Frameset Navigation
-
ParcelYogi Landing Page Clone
-
Responsive Product Comparison Page
Note
Java Automation: Experienced in building Java-based automation tasks on Linux platforms and managing remote deployments/file transfers using WinSCP.


