I design and ship production AI systems for real businesses — webhook-driven pipelines, multi-agent content and publishing tools, and data intelligence that runs on its own. No slideware, things that actually run.
A prototype built on the NxVET ambient voice platform. When a conversation completes, a single webhook pipeline pulls the transcript and SOAP note once and fans it out to three lenses — no staff effort, no app to tap.
One-command publishing of client proposals, ROI calculators, and the marketing site to Cloudflare Pages — with auto Stripe payment buttons and Slack notifications wired in.
A multi-agent Claude system that ships marketing assets across social, blog, email, and Meta Ads through an ingest → preflight → preview → publish pipeline with a mandatory approval gate.
A nine-sub-agent content team under an orchestrator, with a python-docx document generator and two custom MCP image-generation tools.
Turns a discovery-call transcript into a client-tailored sell sheet and pitch deck by filling tokenized Google Slides templates — fixed pricing and brand copy preserved by design.
Hourly poller pulls new recordings, a Claude Opus structured summarizer parses them, and a morning email digest goes out — running unattended on a schedule.
A weekly procurement scraper that scans 95 public portals, scores openings with tiered keyword matching, and emails a relevance-ranked digest. Runs on a Monday cron.
A daily Slack briefing over the Meta Ads API — anomaly flagging on spend and performance, generated headlessly by a scheduled Claude CLI job.
A Slack Bolt bot: /task routes to Airtable across four bases, /schedule creates Google Calendar events with Meet links via a service account.
A productized six-module training curriculum authored in three coordinated formats (learner content, voiceover, AV), delivered through a React 19 + Vite LMS front end.
A 15+ page marketing site rebuild with a structured SEO / answer-engine program: JSON-LD schema suite, llms.txt, and a discovery baseline off Search Console.
An end-of-day scanner combining yfinance price data, FinBERT news sentiment, and Finnhub fundamentals into a dashboard with Slack alerts and a control-group backtest.
An interactive React + TypeScript app for an engineering firm, mapping their sales organization — roles, business units, and a multi-level reporting hierarchy — taken from v1 to v2.
A six-page marketing website on a bespoke brand system, plus a deployable build and pitch collateral, for a co-op network placing AI-skilled students with companies.
Hand-coded, brand-tailored interactive proposals with built-in ROI logic and print-ready PDF export — delivered for Canada Technical Solutions (custom trailers) and the Region of Peel (regional government).
Per-prospect, single-file discovery apps — one question at a time, notes auto-saved, end-of-call summary — used live on sales calls. Built for Nebsys Digital, ApexPower, and Vistage.
Event-driven pipelines that ingest data, run it through AI, and take a real action — bookings, messages, alerts.
Orchestrated Claude agents with approval gates, audit logs, and safety controls that ship to live channels.
Scrapers, scanners, and summarizers that turn noisy sources into a ranked, scheduled digest you actually read.
Hand-coded, brand-compliant sites and interactive tools, wired to one-command Cloudflare publishing.
AI Solutions & Automations Developer at Champion AI.
I'm the person a business hands its "someone should really automate this" list to. I build the automation layer that runs in the background — event-driven pipelines, multi-agent systems, and data tools that quietly do the work while everyone sleeps.
My rule is simple: it has to actually run, not just demo well. Most of what I ship lives on an always-on server, wakes up on a schedule, and never asks for attention. Give me a messy manual workflow and I'll hand back a system that does it on its own — a signal comes in, AI makes sense of it, and something useful happens on the other side.