# Prafull Salunke

Hi, I am Prafull Salunke. This is my professional portfolio website.
I am a Senior Principal Engineer and Architect based in Bangalore, India. I specialize in designing platforms, building scalable infrastructure, and developing sophisticated web architectures, including modern AI agents and systems.

## Sections
- [Home](https://www.prafulls.me/): Overview and latest featured writing.
- [About](https://www.prafulls.me/about): More detailed background, skillset overview, and timeline.
- [Portfolio](https://www.prafulls.me/portfolio): Professional experience, technical skills, and architecture roles.
- [Blogs](https://www.prafulls.me/blogs): Read my articles on engineering, architecture, scaling, and Artificial Intelligence (AI).
- [Contact](https://www.prafulls.me/contact): Get in touch.

## Latest Articles
- [Agentic AI: When Your AI Stops Asking and Starts Doing](https://www.prafulls.me/blogs/agentic-ai-autonomous-systems): What makes AI agents actually agentic, the architecture behind agent loops, and why the trust/control tradeoff is the real engineering problem.
- [Hybrid AI: When LLMs Meet Expert Systems](https://www.prafulls.me/blogs/hybrid-ai-architectures): Why pure LLMs fail at rule-based reasoning, and how hybrid architectures combine language models with deterministic systems to get reliable results.
- [Federated Learning: Training AI Without Seeing the Data](https://www.prafulls.me/blogs/federated-learning-privacy): How federated learning trains models across decentralized devices without moving the data, and the practical problems that come with it.
- [Synthetic Data: Training AI on Data That Doesn't Exist](https://www.prafulls.me/blogs/synthetic-data-generation): How synthetic data solves data scarcity and privacy problems in AI training -- generation methods, quality metrics, real use cases, and where it falls apart.
- [Modular AI: Building Systems You Can Actually Maintain](https://www.prafulls.me/blogs/modular-ai-systems): Designing AI systems with interchangeable components -- module taxonomy, interface contracts, hot-swapping models, and real tradeoffs vs monolithic deployments.
- [AI Behind APIs: Microservices Patterns for ML Systems](https://www.prafulls.me/blogs/microservices-ai-integration): How to expose AI capabilities through microservices -- model-as-a-service, async inference, batching, sidecar patterns, and when a monolith is the better call.
- [MCP in Practice: Building Real AI Integrations](https://www.prafulls.me/blogs/mcp-ai-integrations): A practical implementation guide for building MCP servers -- tool schemas, auth, streaming, testing, and deployment patterns.
- [Multi-Agent Systems: When One AI Isn't Enough](https://www.prafulls.me/blogs/multi-agent-systems): Designing systems where multiple specialized AI agents collaborate -- orchestration patterns, communication protocols, and failure modes.
- [Event-Driven AI: Real-Time Intelligence at Scale](https://www.prafulls.me/blogs/event-driven-ai-architectures): Processing real-time data streams for AI inference -- architecture patterns, feature consistency, backpressure, and graceful degradation.
- [Enterprise Architecture in the AI Era](https://www.prafulls.me/blogs/enterprise-architects-ai-era): How AI is changing the enterprise architect role -- from integration patterns to AI governance, model lifecycle management, and the new enterprise AI stack.
- [Spec-Driven Development: The GSD Way](https://www.prafulls.me/blogs/gsd-spec-driven-development): Why writing specs before code produces better software -- the GSD methodology for solo developers and AI.
- [RAG Without Vector Search](https://www.prafulls.me/blogs/pageindex-vectorless-rag): PageIndex replaces vector search with LLM reasoning over a hierarchical document index.
- [Inside Claude's Agent System](https://www.prafulls.me/blogs/claude-agent-team-architecture): How Claude Code's agent team works -- orchestration, subagent spawning, and context window management.
- [When AI Agents Socialize](https://www.prafulls.me/blogs/ai-agents-socialization): Moltbook, a social network where 770,000+ AI agents formed religions, governments, and encrypted channels.
- [Spark vs Hadoop Showdown](https://www.prafulls.me/blogs/spark-vs-hadoop): Architecture differences, performance tradeoffs, and why Spark has mostly won for new projects.
- [Is AI a Bubble?](https://www.prafulls.me/blogs/is-ai-a-bubble): Market data, historical parallels, and practical advice for engineers navigating the hype cycle.
- [LangChain vs LangGraph Compared](https://www.prafulls.me/blogs/langchain-vs-langgraph): When to use each, architectural differences, and practical patterns for production LLM apps.
- [Kafka Meets Microservices](https://www.prafulls.me/blogs/kafka-microservices-architecture): Event-driven patterns, order processing flows, and production lessons from scaling e-commerce.
- [Designing WhatsApp at Scale](https://www.prafulls.me/blogs/scalable-messaging-platform): WhatsApp-scale messaging platform architecture, message delivery flow, and cost analysis.
- [Local RAG with Ollama](https://www.prafulls.me/blogs/local-rag-ollama): Fully local RAG server using Ollama and ChromaDB -- no API keys required.
- [When RAG Gets Smart](https://www.prafulls.me/blogs/agentic-rag): How autonomous agents improve RAG systems with dynamic, multi-step reasoning.
- [How A2A Protocol Works](https://www.prafulls.me/blogs/a2a-protocol): Google's Agent-to-Agent protocol for multi-agent systems and agent interoperability.
- [MCP: USB for AI Tools](https://www.prafulls.me/blogs/mcp-introduction): Anthropic's Model Context Protocol -- a standard interface for AI tool integrations.
- [Vibe Coding: When Developers Stop Typing and Start Talking](https://www.prafulls.me/blogs/vibe-coding): What vibe coding actually is, why natural language prompts are replacing boilerplate, and where this falls apart in production.
- [AIOps: When Your Pipeline Starts Fixing Itself](https://www.prafulls.me/blogs/ai-augmented-devops): How AI is changing DevOps from reactive firefighting to predictive maintenance, and why self-healing systems are closer than you think.
- [AI-Driven Design Automation: How Machines Are Learning to Design Chips](https://www.prafulls.me/blogs/ai-driven-design-automation): Inside the shift from manual chip design to AI-optimized EDA, what reinforcement learning brings to the table, and why this matters beyond semiconductors.
- [Intelligent Automation: What Happens When RPA Gets a Brain](https://www.prafulls.me/blogs/intelligent-automation-rpa-ai): The evolution from scripted bots to AI-powered automation, where intelligent automation actually delivers ROI, and why most implementations still fail.
- [Serverless Computing: The Promise, the Reality, and the Bill](https://www.prafulls.me/blogs/serverless-computing-guide): A frank look at serverless in 2026 -- what it's actually good for, the cold start problem that won't die, and how to avoid the bill shock.
- [Green Software Engineering: Writing Code That Doesn't Cost the Earth](https://www.prafulls.me/blogs/green-software-engineering): Why software has a carbon footprint, what green software engineering actually means in practice, and concrete steps developers can take.
- [Claude Mythos: The Model Too Dangerous to Ship](https://www.prafulls.me/blogs/claude-mythos-preview): Anthropic built its most capable model ever, then decided nobody should use it. What Mythos can do, why it's locked behind Project Glasswing, and what this means for everyone building with AI.
- [APIs From the Ground Up: Design, Protocols, and Security](https://www.prafulls.me/blogs/api-design-security-architecture): A working reference for API design decisions -- REST vs GraphQL vs gRPC, transport protocols, auth strategies, and the security layers that keep things from falling apart.
- [The One-Prompt Illusion: Why AI Makes Software Look Easy](https://www.prafulls.me/blogs/one-prompt-illusion): Product teams think one prompt builds an app. Engineers know it takes a hundred. The growing gap between AI-generated demos and production-grade software, and how to bridge it.
- [Testing Auth: What Actually Breaks in Production](https://www.prafulls.me/blogs/auth-testing-best-practices): A practical guide to testing authentication systems, from JWT edge cases to OAuth callback bugs and session race conditions that only show up at scale.

## Topics
AI Agents, Agentic AI, RAG, Retrieval-Augmented Generation, LLM, Model Context Protocol, A2A Protocol, Spec-Driven Development, Developer Productivity, System Design, Microservices, Event-Driven Architecture, Apache Kafka, Distributed Systems, Cloud-Native, Software Architecture, Hybrid AI, Expert Systems, Neuro-Symbolic AI, Federated Learning, Edge AI, Synthetic Data, Modular AI Systems, MLOps, Multi-Agent Systems, Enterprise Architecture, AI Governance, Vibe Coding, AIOps, DevOps, CI/CD, Observability, Design Automation, EDA, Chip Design, Reinforcement Learning, Semiconductors, Intelligent Automation, RPA, Process Automation, Serverless, Cloud Computing, AWS Lambda, Vercel Functions, Green Software, Sustainability, Carbon Footprint, Claude Mythos, Project Glasswing, AI Safety, Cybersecurity, Anthropic, API Design, REST, GraphQL, gRPC, OAuth, JWT, Authentication, Authorization, WebSockets, CORS, Authentication Testing, Security Testing, Session Management, Token Security