From support tickets to production AI.
In under a year.
A year ago, I knew some basic JavaScript and worked in customer support. Today I build production AI systems — full-stack, deployed, award-winning — without writing code manually. I architect, orchestrate, test, and make decisions. AI handles the syntax. This portfolio is the proof.
Agency Intelligence Engine
An AI system that discovers, profiles, and tracks e-commerce agency partners from support data — turning invisible signals into actionable intelligence.
A 70-person Shopify app startup had thousands of support conversations happening daily. Buried inside them were high-value signals about agency partners — but nobody was capturing them. The data was invisible.
A full-stack AI system with 7 agents in an automated pipeline. One classifies incoming data, another scans e-commerce stores for agency footprints, another builds full agency profiles using a multi-round AI pipeline, and another drafts personalized outreach emails. Three supporting agents handle deduplication, orchestration, and metrics.
Used Claude Code to run multiple parallel development sessions — backend, frontend, database, AI pipeline — simultaneously. Built in ~2 weeks. I was the architect and decision-maker. AI wrote the code. I made sure it was right.
Pioneered automation infrastructure at a 70-person startup — connecting the entire tech stack before anyone else.
Connected 9 services into a central automation platform — support tools, AI APIs, analytics, CRM, workspace tools, partner platforms. None of these connections existed before. When the company later ran a month-long sprint focused on AI and automation, the infrastructure was already there. People could focus on building workflows because the plumbing existed.
I don't write code manually. I architect systems and orchestrate AI to build them.
Built automations for people outside my team — reducing hours of manual work to minutes.
A Chrome Extension that centralizes everything a support agent needs into one interface.
Operates across three contexts: inside the support platform, inside the e-commerce admin panel, and as a standalone popup. Entirely vibe-coded over 2 years. Not a single line of hand-written code.
Discovered that the AI chatbot can call automation workflows mid-conversation — and built the first working example.
The chatbot had no way to take real actions — only answer from a knowledge base. I figured out it could call an external workflow on demand. First example: real-time quota checks without a human agent. Opened up an entire category of automations.
An AI assistant that reads entire support conversations and drafts personalized emails in my writing style.
I give it a conversation ID and my brief notes. It reads the full thread, understands the context, and drafts a ready-to-send email. Personalized and contextual — not generic templates.
Identifying at-risk customers from conversation patterns before they churn. Same pipeline architecture as Night's Watch, different classification rules. Research done. Architecture designed. Planned for future expansion.
Currently exploring new projects. If you're building something interesting or need help with AI and automation, I'd love to hear about it.
CONNECT ON LINKEDIN