Kartik M.

I build and troubleshoot AI agents, voice assistants and business automations using n8n, Claude, OpenAI APIs and Python. My work includes connecting business tools, automating research and reporting, and extending existing applications with useful reporting features. I'm based in Hamburg, Germany, and currently hold Top Rated status with 100% Job Success on Upwork. My completed projects include local AI setup on a Mac mini and voice agent development for an agency serving auto repair shops. For a German employee training platform, my work included protected manager views, paginated exam reports and filtered video activity reports within an existing application. That project uses React, TypeScript and Supabase. The work also covered learning paths, access rules and progress tracking. For another German business client, I maintain and extend an OpenClaw system on a VPS. The workflows use Google Sheets for inputs, Google Docs and Gmail for reports, and APIs and MCP tools for research and publishing. I investigate failed scheduled jobs, check whether outputs reached their destination, and prevent duplicate delivery when a run needs to be retried. I've also installed and debugged OpenClaw and Ollama on a client's Mac mini M4. I tested the setup against its 16GB memory limits and delivered an evaluation report, a custom installation script and a practical client guide. My infrastructure work includes private device access through Tailscale. With n8n, I've built a separate research workflow that sends tasks to Hermes, checks returned fields and source evidence, and saves research and German drafts for review. I work with REST APIs, webhooks, JSON responses and tool calling to connect the steps and trace failures. My Vapi work includes live voice agents for auto repair businesses, caller information capture and appointment request flows. I test conversations and tool responses, including date handling and input validation, so the agent passes useful information to the business. I also develop agent harnesses and tools around Claude Code and OpenAI Codex. My personal Open Harness project explores coordinating coding agents, reviewing their work and managing tool access. My development tools include Python, JavaScript, TypeScript and GitHub. AI evaluation is another part of my work. I review coding responses, instruction following and tool use, write realistic test tasks, and document hallucinations, incomplete actions and repeatable failures. My evaluation work has also covered Make, Zapier and Microsoft 365 workflows. I can help with a new automation, reporting features, an existing agent that needs debugging, or a review of reliability and model behaviour. Send me the workflow, the tools you already use and the part that is causing trouble. I'll help define a practical first step and document the result clearly.