🌐 Portfolio | 💼 LinkedIn | 💻 GitHub | 📧 Email

interface SouvikGhosh {
  role: string = "Full-Stack & AI Developer";
  expertise: string[] = [
    "Java Backend Development",
    "TypeScript & Node.js",
    "Android Development",
    "Flutter Cross-Platform",
    "AI & Machine Learning",
    "Retrieval Augmented Generation (RAG)",
    "AI Agents & MCP Frameworks",
    "React & Web Development"
  ];
  specialization: string[] = [
    "Model Context Protocol (MCP)",
    "Large Language Models (LLMs)",
    "Vector Databases & Embeddings",
    "Incident Automation Systems",
    "REST & Event-Driven APIs",
    "Multi-Agent Orchestration"
  ];
  currentFocus: string = "Building intelligent AI systems, RAG pipelines, AI Agents, and MCP-based automation";
  openTo: string[] = ["Full-time roles", "Freelance", "AI/ML Projects", "MCP Development"];
}

I'm a passionate full-stack developer with deep expertise in Java Backend, TypeScript, Android, Flutter, and cutting-edge AI & RAG Technologies. I specialize in building intelligent systems powered by LLMs, creating sophisticated RAG pipelines, developing autonomous AI agents, and architecting MCP-based automation frameworks. I love solving complex problems and contributing to open-source projects.

• 🔧 Model Context Protocol (MCP) Architecture
• 🤖 Advanced AI Agent Orchestration & Agentic Workflows
• 📊 Vector Database Optimization (Pinecone, Weaviate, FAISS)
• 🧠 Fine-tuning Large Language Models & Prompt Engineering
• 🔗 Multi-Chain RAG Systems with Knowledge Graphs
• 🚀 LangChain, LlamaIndex & Advanced LLM Frameworks
• ⚡ Real-time Event-Driven Architectures
• 🌍 Serverless & Microservices Patterns

• 🔍 Advanced RAG Systems - Building scalable RAG pipelines for enterprise applications
• 📚 Vector Embeddings - Working with OpenAI, Hugging Face embeddings for semantic search
• 🗄️ Vector Databases - Pinecone, Weaviate, FAISS, Chroma integration
• 🔗 Multi-hop Retrieval - Complex document retrieval and reasoning chains
• 📄 Document Processing - PDFs, Web content, Markdown parsing and chunking

• 🧠 Autonomous Agents - Building agents with tool use and function calling
• 🔄 Agent Orchestration - Multi-agent systems with inter-agent communication
• 🎯 Tool Integration - Connecting agents to APIs, databases, and external services
• 💭 Reasoning Chains - Chain-of-thought prompting and reflection loops
• 🚀 Production Deployment - Scalable agent architectures on cloud platforms

• 🔧 Model Context Protocol (MCP) - Building MCP servers and clients for LLM integration
• 🤝 Agent-LLM Communication - Standardized protocols for tool-use and resource access
• 📡 Event-Driven Architecture - Real-time agent coordination and incident automation
• ⚙️ Automation Frameworks - Building intelligent automation systems with MCP

• 🎓 LLM Fine-tuning - QLoRA, LoRA techniques for model customization
• 💬 Prompt Engineering - Advanced techniques for optimal LLM performance
• 🔄 LangChain & LlamaIndex - Building complex LLM applications
• ⚙️ Model Integration - OpenAI, Anthropic, Hugging Face, Ollama
• 📊 Evaluation Metrics - BLEU, ROUGE, METEOR for LLM output assessment

• Java Backend Developer - Building robust, scalable server-side applications
• TypeScript & Node.js - Full-stack development with modern JavaScript ecosystems
• Flutter & Android - Cross-platform & native mobile development
• AI & Machine Learning - LLMs, RAG systems, and AI agents
• Retrieval Augmented Generation - Building intelligent search and Q&A systems
• Model Context Protocol (MCP) - Incident automation and agent frameworks
• AI Agents - Autonomous systems with tool use and multi-agent orchestration
• Full-Stack Development - End-to-end application architecture

A stunning, fully responsive portfolio website
built with modern technologies and best practices.

A classic Tic Tac Toe game built with modern
technologies featuring interactive gameplay
and smooth user experience.

Advanced incident automation system using
Model Context Protocol (MCP) for intelligent
incident management and response automation.

Android app for editing text files and sharing
files seamlessly via WiFi P2P with compiler.

Feature-rich Ludo game application with
interactive UI and smooth real-time gameplay.

Intelligent system optimizing retail inventory
using ML algorithms and Java backend architecture.