AI & Automation

AI that actually does something
useful for your business.

I've gone deep into AI — not as a user of tools, but as someone who builds with the APIs, runs the infrastructure, and ships products on top of it. There's a big difference.

6+ AI providers integrated
in production systems
3 AI products built &
deployed for clients
Local Ollama + open source
models running on-prem
FR+EN Bilingual AI — a Quebec
market differentiator

Most "AI for business" advice is shallow. This isn't.

The internet is full of people who figured out how to use ChatGPT last year and now sell AI consulting. That's not what this is.

I've been working with AI APIs directly — writing code that calls Anthropic, OpenAI, Google, and others, routing requests through Cloudflare AI Gateway for unified billing, building RAG pipelines, handling embeddings, managing token costs, and deploying self-hosted models for clients who can't send data to the cloud.

The difference matters when you're making a real business decision.

A chatbot that answers three questions then breaks
A properly trained chatbot with RAG knowledge base
Prompt templates from a course
Production prompt engineering from real use
One provider, one model, no fallback
Multi-provider routing with automatic failover
Your data going to OpenAI's servers
Self-hosted models when privacy matters
AI that says whatever it wants
Gated, controlled output — enterprise-safe

Things I've actually shipped.

Not demos. Not experiments. Production systems with real users.

SaaS Product

Phasma — Managed AI Chatbot Platform

A full managed AI chatbot service I designed, built, and operate — targeting fitness coaches, studios, and non-profits in Canada. Bilingual English/French out of the box. Clients get a setup fee and a monthly subscription; I handle everything technical.

  • Hub-and-client WordPress plugin architecture — central hub manages all configs, lightweight client plugins on customer sites
  • Multi-provider AI routing: Claude (Anthropic) and GPT-4 (OpenAI), switchable per client
  • Cloudflare AI Gateway integration for unified billing and request logging across all providers
  • RAG pipeline — knowledge base chunking, OpenAI embeddings, SQLite vector search, context injection
  • Voice input via Whisper API transcription
  • Encrypted API key storage, license management, rate limiting, lead capture
  • Full bilingual widget (EN/FR) — differentiator for Quebec market
  • Phasma Deep — membership integration plugin with HMAC-signed per-hour tokens, tier-based access (basic/premium), MemberGate support, hashed member IDs for privacy
  • AI Chatbot Prepper — a separate Python CLI tool that scrapes a client's site and uses Claude to auto-generate their chatbot system prompt
PHP / WordPress

Python

Cloudflare Gateway

Claude + GPT-4

RAG / Embeddings

Whisper Voice API
Personal Tool

AI Trading Dashboard

A personal Streamlit trading dashboard with AI-powered market analysis. Multi-tab interface with candlestick charts, catalyst scanning, backtesting, and an AI morning brief that summarizes market conditions before the open.

  • Cloudflare AI Gateway unified billing — one API token across Anthropic, OpenAI, Google Gemini, Meta Llama, Mistral, DeepSeek
  • Model-agnostic routing layer — swap providers at runtime from the UI
  • Morning brief auto-generates at 6am ET from live RSS feeds + technical snapshots
  • Catalyst scanner scores full watchlist 0–100 across six signals
  • Extended hours data, support/resistance levels, Questrade API integration
Python / Streamlit

6 AI Providers

Cloudflare Gateway

Questrade API
Automation Tool

AI Chatbot Prepper

A Python CLI tool that researches a business from a URL, scrapes and extracts key information using Claude, then auto-generates a production-ready chatbot system prompt and gateway import config. What used to take an hour now takes minutes.

  • Sitemap-aware web scraper with fallback spidering
  • Claude-powered structured data extraction (services, hours, FAQs, tone, audience)
  • Interactive gap-filling — asks targeted questions for missing fields
  • Outputs a polished system prompt + 59-column CSV for direct gateway import
Python

Claude API

BeautifulSoup

I work directly with the APIs — not just the apps.

Most people use AI through ChatGPT or Claude.ai. I work with the underlying APIs directly — writing code that calls them, handling authentication, managing token limits, normalizing responses across different provider formats, and routing traffic intelligently.

Through Cloudflare AI Gateway, I've built systems that route to Anthropic, OpenAI, Google, Meta, Mistral, and DeepSeek from a single API token — with unified billing, automatic logging, and the ability to switch models or providers without touching client infrastructure.

What this means for you: You're not locked into one provider. If OpenAI goes down, the system routes to Claude. If costs spike, I can switch to a cheaper model. If your data is sensitive, I can point it at a self-hosted model. That flexibility is only possible when someone understands the full stack.

Commercial APIs — direct integration

Anthropic / Claude OpenAI / GPT-4 Google Gemini Cloudflare Workers AI

Via Cloudflare AI Gateway

Meta Llama Mistral DeepSeek Unified billing Auto-failover

Embeddings & RAG

OpenAI text-embedding-3 SQLite vector search Chunk + retrieve pipelines

Voice

OpenAI Whisper API Voice-to-text in production

Prompt engineering

System prompt design RAG context injection Structured output / JSON mode Multi-turn conversation management

When your data can't leave your building.

Healthcare providers, legal firms, HR departments, and anyone handling sensitive client data has a real problem with cloud AI — their information goes to someone else's server.

I run Ollama locally with open source models (Llama, Mistral, and others) and deploy OpenWebUI for clients who need a private, self-hosted AI interface. Same ChatGPT-style experience, zero data leaving your infrastructure.

This isn't a theoretical capability. I run it. I can set it up for you.

🦙

Ollama

Local LLM inference on your own hardware. Running Llama, Mistral, and other open source models. No API calls, no data leaving the machine.

🖥️

OpenWebUI

Self-hosted ChatGPT-style interface I deploy for clients. Your team gets a familiar AI chat experience on your own server, connected to your models.

🔒

Private deployment

Runs on your VPS or dedicated server. SSL secured, access controlled. No subscription to a third-party AI service required.

What I can do for you.

Concrete engagements, not vague "AI strategy."

💬

AI Chatbot — Managed Service

A properly configured, bilingual chatbot for your website through Phasma — my own platform built for fitness coaches and non-profits. Setup fee plus monthly. I handle everything.

📚

RAG & Knowledge Base Training

Give your AI actual knowledge about your business — products, policies, FAQs, procedures. I build RAG pipelines that chunk, embed, and retrieve your content so the AI answers from your data, not guesses.

🔌

AI API Integration

Adding AI capabilities to your existing systems — WordPress, WooCommerce, ActiveCampaign, custom apps. I write the integration code, not just configure a plugin.

🏠

Private AI Deployment

Ollama + OpenWebUI on your own server. Your team gets a ChatGPT-like tool, your data stays on-premises. Ideal for healthcare, legal, HR, or anyone with sensitive client information.

🛡️

Safe & Enterprise-Ready AI

AI that stays on-topic and on-brand. I implement stop words, output gating, topic restrictions, and guardrails so your AI doesn't go off-script, embarrass you, or answer things it shouldn't. Important for any client-facing deployment.

⚙️

AI Workflow Automation

Connecting AI to your existing business processes — content pipelines, lead qualification, customer communication, report generation. Real automation, not demos.

🎓

AI Training for You & Your Team

Practical, hands-on training in how to get dramatically more out of the AI tools you already pay for. Not a generic course — tailored to your business, your tools, and your actual workflows. I'll show your staff what's possible and how to get there.

🧠

Prompt Engineering & LLM Tuning

System prompt design, context management, temperature and parameter tuning, structured output, multi-turn conversation handling. Getting the model to behave exactly the way your use case requires.

🇨🇦

Bilingual AI for Quebec

French and English AI systems that actually work in both languages — not translated, natively bilingual. A real differentiator if you serve the Quebec market.

Let's talk about what AI can actually do for your business.

Not a pitch deck. A real conversation about your situation and whether there's a genuine use case worth building.

Start the conversation