📍 Bangalore, India | 🧰 Full-stack & AI platform engineer | ✈️ 3 years at Amadeus

I build the boring, load-bearing parts: Java and Python backends, the Kubernetes and Helm underneath them, and lately the agents that test, file issues and raise pull requests on their own.

• 🏛️ MoG-India - the Union Government of India as a searchable, zoomable graph of 370 bodies (live demo)
• 📄 planview - local MDX doc viewer: browse, comment, export, nothing leaves your disk (npx @vtripathi/planview)
• 🚦 risk-gate - deterministic risk-based code review gates that auto-merge the safe PRs
• 💬 v_slash - mention it in Slack, it reads your GitHub repos, answers, and opens small PRs
• 🧠 skills - agent-agnostic skills for AI coding assistants

• 🧠 skills - visual-docs, design-de-slop, copilot-optimize, retro-gate, design-doc, plus a curated SDLC subset of pstack and compound-engineering
• 🚦 risk-gate - P0/P1/P2 path tiers as a GitHub Action, schema/docs/coverage/lint/plan-link/handoff gates, a bot-noise meter and a Copilot review mesh with no LLM key
• 💬 v_slash - personal Slash-style Slack + GitHub assistant on Slack Bolt, PyGithub and the Anthropic SDK
• 📄 planview - Next.js viewer for .mdx plans with inline notes, git history and HTML/PDF export; pairs with the visual-docs skill

• 🏛️ MoG-India - React 18 + Cytoscape visualization of ministries, departments, regulators, tribunals and PSUs, with cited constitutional powers for every element
• 🍽️ FoodFinder - React + Express + MongoDB app that compares restaurant menu prices from Swiggy, with GPS lookup and a TTL cache

• 🧩 neetcode-submissions - NeetCode solutions in Python, synced automatically

• 🧭 career-ops - ported from Claude Code to GitHub Copilot and kept in sync with upstream
• 🧬 cognee - memory control plane for AI agents
• ☸️ minikube - local Kubernetes
• 🧩 cursor/plugins - Cursor plugin specification and official plugins
• 📚 ai-primitives-hub - Amadeus prompt and skill registry for Copilot

• Closing the loop - QA agents that run scenarios and file issues, and builder agents that fix them and raise PRs, with humans only starting the session
• Modernising old Java - JDK 17, JBoss EAP 8, Podman, Helm; the migrations nobody tweets about
• Packaging what I learn as skills - so the next agent (or person) starts further ahead

• Wrapping up 3 years at Amadeus in October 2026 and looking for the next full-stack, platform or AI-agent engineering role
• Keeping the open-source tools above alive - skills, risk-gate, planview and v_slash
• Exploring the agent loop as a pattern outside any one company's pipeline
• Grinding problems - on LeetCode and NeetCode

"Own the unglamorous parts." Most of the value I've added came from migrations, connectivity descriptors and local dev setups that let other people ship. The AI agents are new; the habit isn't.