I build agentic AI systems — orchestrating multi-agent workflows with guardrails, security, and intentional design. 17+ years turning complex problems into elegant software.
With 17+ years of full-stack engineering under the belt, I'm now channeling that depth into the frontier of AI — specifically, building agentic systems that actually work in production.
I'm not just calling APIs. I'm architecting controlled, multi-agent workflows with proper guardrails, security boundaries, and feedback loops — the kind of systems that can be handed off to a real engineering team.
Every project I build teaches me something new about where AI agents break down and how to make them robust. I believe the next generation of software isn't just AI-assisted — it's AI-orchestrated.
Each project is a real system, not a tutorial. Built with intention.
A complete multi-agent pipeline that takes a user story for any GitHub repo and autonomously plans, codes, reviews, and runs CI/CD checks — finally submitting a real PR ready to merge.
An AI-driven PWA built for kids. It greets children by name, listens to their story request, then generates a fully narrated, illustrated bedtime story — personalised in real time.
A classic 6-player South Indian partnership card game, rebuilt as an online PWA during COVID. Features a full rules engine and AI bots with human-like gameplay — so you can play solo or with friends.
An intelligent automation system that orchestrates team selection workflows with persistent backend scheduling, skill-based balancing algorithms, and state management — overcoming cold start challenges and ephemeral storage limitations for reliable, continuous operation.
An agentic PWA that autonomously tracks product prices across retailers using deterministic-first multi-agent intelligence — reducing LLM costs while delivering proactive deal alerts when prices hit target thresholds.
Enterprise MCP-powered workflow orchestration system that automates the full incident-response lifecycle across variosu plaforms(ServiceNow, Jira, Microsoft Teams etc.) - featuring a LangGraph state machine, human-in-the-loop approval gates, real-time Angular dashboard via SSE, PostgreSQL persistence, and a full Docker stack.
Autonomous developer agent platform that takes a Jira story end-to-end - breaking it into tasks, then autonomously planning, coding, reviewing, and running CI/CD - before submitting a PR for human approval. All agent-generated code executes inside an isolated Docker sandbox with no network access, token budget governance, circuit breakres, and prompt injection sanitization at every boundary.
Enterprise-grade RAG system that acts like "Google for your confluence" - employees ask natural-langugage questions and recieve accurate, cited answers synthesized from internal documentation. Built with FAISS vector store, LangGraph orchestration for ambuiguity detection and milti-turn clarification, an Angular answer, panel with source cards, a feedback loop, and an analytics dashboard with full Docker deployment.
Paste a job description. My AI agent tailors a resume, scores skill match, and flags gaps — in seconds.
This isn't a chatbot that answers FAQs. It's an agent that flips the dynamic — it asks smart questions about your company, your engineering challenges, and what you're hiring for. Then it tells you exactly how Praveen maps to your needs.