Projects · Section 01
Working systems — a trigger, automated steps, decision logic, and a reliable result at the end. Each one is a small pipeline built and tested end to end.
A reservation landing page and CRM automation built inside GoHighLevel — the platform most in-demand right now among agencies and local businesses. A Tally form submission gets checked against existing open reservations, routes into a Found/Not Found branch, and either notifies staff of a possible duplicate or creates a new opportunity and confirms it.
The real work was deliberate destroy-testing — deliberately trying to break the automation before trusting it with anything real. That surfaced three non-obvious bugs (a misleading action label, a stale-match problem with repeat customers, and a contact-deduplication gap that let typo'd emails slip through), each root-caused and fixed at the layer where it actually lived.
A fully deployed conversational AI agent running on this portfolio — answering visitor questions 24/7, detecting hiring interest through natural conversation, and triggering a 10-step automated pipeline the moment a lead is qualified. No forms. No manual work.
When a visitor shows interest, the agent collects their name, project, and email through conversation. A pipeline fires automatically — logging the lead to Google Sheets, sending an instant owner alert, and delivering a personalized follow-up email to the lead — all before the chat ends. Built entirely on a free stack: Groq AI, Cloudflare Workers, and Google Apps Script.
A fully deployed conversational AI agent running on the Automate with Bon Facebook Page — answering Messenger messages 24/7, responding naturally as Bon, qualifying leads automatically, and triggering a complete pipeline without any manual work.
When someone messages the page, the agent responds instantly as Bon — answering questions about services, background, and availability. When hiring interest is detected, it collects name, project, and email naturally through conversation, logs the lead to Google Sheets, and sends an automated Gmail notification. Built on a free stack: Make.com, Groq AI, and the Meta Messenger API. Live and publicly accessible.
A second Messenger AI chatbot, rebuilt from scratch on Activepieces after shipping the first one on Make.com — same core architecture, different platform, built deliberately to get hands-on with how a different automation builder handles webhook verification, expression syntax, and step-level testing.
The flow catches every incoming Meta webhook event, routes verification handshakes separately from real messages, checks for duplicate fires, pulls conversation history for context, sends the message to Groq AI for a reply, logs the exchange with a live timestamp, and sends the reply back via the Messenger Send API. Fully wired end to end — currently gated by Meta App Review, so it's not public yet.
A fully automated job-search pipeline that scans remote job boards daily, filters genuine matches against target keywords, and for every new match, generates a resume and cover letter tailored to that exact posting — not a generic template, rewritten every time. Delivered straight to email each morning, with the original job link included.
Sources only from job boards with free, public APIs — no scraping, no auto-submitting. It prepares every application; a human still sends it. Scheduled via GitHub Actions, running on $0/month infrastructure end to end.
A fully deployed appointment booking agent that takes a Tally form submission, checks real availability on Google Calendar, books the slot or asks AI to propose alternatives when it's taken, logs every attempt to Google Sheets, and sends a confirmation email — all without any manual work.
Two distinct outcomes, not one: a free slot books automatically, while a conflict hands the actual overlapping events to Google Gemini AI to write a natural reply. Every attempt is logged, whether confirmed or conflicted.
A fully deployed review response agent that takes a Tally review submission, validates it, and routes it by star rating into three distinct branches — negative, neutral, and positive — each handled with the exact mix of AI and templated logic that rating actually needs.
Negative reviews get a Google Gemini AI-drafted apology with a discount code, but it's held for owner approval before ever reaching the customer. Neutral reviews get a lighter AI reply sent directly. Positive reviews skip AI entirely — a templated thank-you is all that's needed. Every review is logged to Google Sheets with its sentiment and status.
Filters incoming emails to catch new orders, uses AI to extract order details from unstructured text, matches items against a pricing database, and calculates totals and profit — then logs everything to Google Sheets, follows up with a Tally feedback form, applies AI sentiment analysis to generate personalized responses based on the customer's rating, and sends a personalized reorder invitation 24 hours later.
A fully automated content production system that runs on a daily schedule — using AI to generate scripts, JavaScript to process and format the output, Visla to create video projects from the script, and Google AI Studio (Gemini) to generate thumbnails.
From there, the pipeline cross-posts automatically to YouTube and Instagram for Business, logs every publish to Google Sheets, and sends email notifications at each stage — with conditional path logic handling success, fallback, and error states throughout the entire flow.
A fully automated client pipeline built inside Trello — moving cards across 7 stages, swapping checklists per stage, auto-assigning due dates, flagging urgent work, and sending email alerts — all triggered by zero manual input after the initial setup.
When a new client card is created, the automation fires immediately — setting a due date, description, checklist, and email notification in one rule. As the client progresses, the checklist swaps automatically, labels update, an audit comment is logged with the date, and the due date resets for the next stage. Built entirely on Trello's free Butler automation — no external tools, no paid plans, no code required.
Triggers every day at 7 AM, checks the forecast, uses AI to turn it into a short readable summary, and sends it straight to an inbox — a small example of recurring reports that run themselves.
This list grows as new automations get built and tested. Check back soon, or get in touch if there's a specific workflow you'd like built.