About Me
Hi, I'm Kartik. Senior Frontend Engineer, 10+ years, based in Amsterdam.
Engineering with AI agents
Most of my recent curiosity has gone here. On a side-project (a small SaaS I'm building from scratch), I moved past the "Claude as autocomplete" phase and started designing the workflow itself.
I built a custom MCP server (in Go) that exposes the backend's schema (endpoints, types, auth) to the frontend agents in a sibling repo, so the AI on one side can introspect the other without reading source code.
On top of that, I wrote 10 Claude Code skills that orchestrate work as a pipeline:
orient > brainstorm > groom > execute > spawn-epic
A PreToolUse hook blocks commits when the phase gate
(build, test, lint) fails, and 13 per-module CLAUDE.md
files auto-load whenever an agent descends into the right folder.
The whole thing runs on a stateless context-engineering pattern I designed: state lives in the filesystem (folder = state), logs are append-only, and a two-layer constitution (workspace plus epic) keeps multiple initiatives from drifting. I'm writing it up as a long-form blog post.
At work I apply the same playbook: GitLab MR review automation, agent tools, custom skills, and hooks for internal workflows.
Tech I lean on
Angular, TypeScript, RxJS, NgRx, Nx monorepo, Web Components, SCSS. PostgreSQL and Go on the side-project side.
People side
I completed Google's People Management Essentials specialization on Coursera (SMART goals, coaching, feedback, building team culture). Across roles I've mentored juniors and interns, owned onboarding for new hires, run code reviews, and prioritized P0 defects across teams. The people work matters to me as much as the code.
Recent learning
Anthropic's Claude Code in Action: tool integration, context management, MCP servers, custom hooks, and GitHub workflow automation. Recommended if you're shipping code with AI assistants daily.