I build AI automations: AI agents and workflows for real business tasks. I make them with Claude Code and connect them to your tools through n8n and API integrations.
My AI DM Setter demo qualifies a lead in chat and books a call with no calendar link. The model replies in strict JSON, so code decides the booking. One workflow can serve many businesses, and I tested that two of them share no lead history. Broken replies get retried, and a fallback still answers the lead.
In agent-os, a manager agent plans work for ten worker roles over a SQLite queue. Model calls fail over across Groq, OpenRouter and local models. Its public repo has 55 tests.
Claude Code is how I build, and I direct it. On my chase-ladder repo, Claude Code writes tests first, and I approve every diff and do the merges. It has core logic and a HubSpot API client, with 176 passing tests. It doesn't send
On the DM Setter, a coding agent used an outdated database setting. I found it in the execution traces through n8n's API and patched it through that API. Coding agents make confident mistakes, so I build a way to see what ran.
I self-host n8n in Docker. I also run and update n8n on a Linux VPS. I build workflows as JSON files and push, patch and audit them through the n8n API, with no secret in the file. I prove failure paths by breaking them on purpose, like pointing a mail step at a dead port.
For a client, I built an n8n lead automation. A form submission
I call Claude's API from n8n with a strict output schema, and build webhooks that need a secret header.
I have a Computer Engineering degree and write Python, C, JavaScript and SQL. I'm HubSpot Academy certified in Inbound and
Message me with one job your team repeats every week, and I'll show you how I'd automate it.
Experience: 2 - 5 years
Experience: 1 - 2 years
Built and ran AI automation end to end: AI DM Setter on self-hosted n8n and Postgres serving many client accounts from one workflow, and a Job AI Agent that cut my own 2 to 4 hour daily task to under 10 minutes. Repos and a walkthrough are linked in my description.
Experience: 6 months - 1 year
Experience: 1 - 2 years
Ship LLM systems that hold up in production: JSON output contracts, targeted retries, history-aware fallbacks, local models on Ollama, RAG with ChromaDB, and a prompt-injection detector reported honestly (tuned F1 1.000, held-out recall 0.350). CpE thesis: MobileNet CNN at over 90% validation accuracy running on a Raspberry Pi 5.
Experience: Less than 6 months
I write against the HubSpot API directly. Built the client for a quote follow-up system on HubSpot Free: it reads the deal pipeline, maps stage labels to stage ids so renaming a stage in the portal breaks nothing, finds due deals and patches them, and backs off on a 429. Tested against an injected fake transport, so it runs with no network.
Experience: Less than 6 months
One site in depth over the WordPress REST API on a paid automation assessment. Built a tool that turns a variable sheet and a content package into finished pages: blocks, reusable elements, menus and Rank Math meta through Application Passwords. It proposes every change in a dry run first, and refuses to write when the live page is not what the plan expected.
Experience: 1 - 2 years
Built the Job AI Agent dashboard and its REST endpoints on FastAPI. RoScript AI Pro routes requests across several LLM providers with failover and key rotation.
Experience: 1 - 2 years
Job AI Agent runs Playwright scraping in parallel, then scores every listing with an LLM for fit and scam signals. Playwright means pages that render with JavaScript are fine.
Experience: 1 - 2 years
I design prompts as contracts: the model must return reply, booked and slot as JSON, and a targeted retry took contract failures from one turn in five to zero. Also built prompt-sentinel, a prompt-injection detector with a Python and TypeScript parity test.
Experience: 1 - 2 years
AI DM Setter is a chatbot on n8n that answers leads and books the call inside the conversation, one workflow serving many client accounts with proven per-client isolation. Also a fully local assistant on Ollama with RAG over notes.
“The more I stepped away from it, the more successful our Chanel became!”
- Jim Orr
Onlinejobs.ph "ID Proof" indicates if "they are who they say they are".
It DOES NOT indicate skill level.
ID Proof scores are 0 - 99 with 99 being the best. It is calculated based on dozens of data points.
It's intended to help employers know who they're talking to is real, and not a fake identity.