15+ years of technical delivery across AI systems, backend engineering, and cloud automation. LLM architectures, multimodal generation, and end-to-end AI pipelines — shipped in production.
AI Engineer and Workflow Automation Consultant with 15+ years of technical delivery across AI systems, agentic coding, and cloud automation.
Specialises in LLM architectures, Claude Code / multi-agent workflows, and full-stack AI products — Next.js on the front, Supabase + MongoDB on the back, Vercel on the edge.
Ships 13 production sites, 34 marketing pages, and a fleet of managed agents — all orchestrated from a single cockpit built with the same stack I sell.
What I reach for in 2026 — Claude Code for agentic build, Next.js + Vercel for shipping fast, Supabase + MongoDB for data.
Three builds that illustrate the stack: an orchestration cockpit, an agent chain, and a daily content engine — all running in production.
Next.js 15 orchestration cockpit for a fleet of 13 production sites. PBI → build → preview → push → deploy, all from one interface. JSON-on-disk registry, Supabase auth, Vercel-survivable read-only mode.
13 sites live, all on the same push-to-Vercel pipeline, managed from a single cockpit.
Discovery → Strategy → Design → Engineering → Distribution, each a Claude Code skill in production. IntakeAgent turns a voice memo into a Project Card; ProposalAgent outputs a full proposal pack in one POM.
Every client engagement runs through the chain — median time from brief to proposal: < 1 POM-day.
Five daily workflows (page, list, image, audio, video) running on a cron via AIOS. Each generates, archives, and upserts to MongoDB AIDB.DAILY. ≈810 top-100 lists in the archive, 200+ hours of audio, 1,000+ videos.
Six content streams running daily — one human touch point per stream (approval).
LLMs · ML / DL · Tokenization · Prompt & Context Engineering · AI memory & grounding · Data integration · Function Calling · Fine-Tuning
Embeddings · Vector DBs · Semantic & Hybrid Search · Chunking · Knowledge Graphs · Retrieval Optimisation · LlamaIndex · Assistants API
Multi-agent orchestration (n8n) · Agentic Planning · Tool Use · Memory / State Mgmt · Voice / Research Agents · AutoGPT / BabyAGI flows · MCP
TTI (SDXL, Fooocus) · TTV (Runware, Sora, Cloudware) · TTS / STT (Whisper, Coqui) · Multimodal pipelines
GPT-5 · Claude · Gemini 2.5 · Hugging Face · Open-source LLMs · OpenAI · Google AI Studio · OpenRouter · Runware · Alibaba Model Studio
Python · FastAPI · Streamlit · REST / JSON APIs · Inference pipelines
AWS / GCP · EC2 · S3 · Docker · CI/CD · NGINX
MongoDB schema design · Data Pipelines · Metadata automation · Vector Search · Statistics · Linear Algebra · Calculus
React · Tailwind · shadcn/ui · Bootstrap · Jinja2 · Rapid prototyping
Agile / Scrum · MVP delivery · Lean · User Stories · JIRA · MkDocs · Confluence · GDPR · Responsible AI · Safety Testing · Explainability · Cost Optimisation · Storytelling · Stakeholder Alignment
From digital transformation to AI engineering — shipping production systems for regulated, large-scale organisations.
Building and shipping production AI systems: 13 live sites on a Claude Code–orchestrated fleet, a 19-agent ABC squad, daily content pipelines (PAD · LAD · IAD · AAD · VAD), and the APPAI cockpit. Client work spans LLM apps, RAG pipelines, and managed-agent deployments.
Delivered Copilot Studio assistants, voice automation, and AI-driven contact-centre workflows.
Integrated AI translation workflows and led Agile delivery across engineering teams.
Delivered CI/CD data pipeline (Snowflake + Jenkins) for marketing analytics automation.
Delivered integrated e-commerce + leisure management platform (membership, payments, ticketing).
Delivered e-commerce booking platform with AI-ready content, search, and payments architecture.
MongoDB University · Snowflake University · Coursera
Concordia University, Montreal
Collège Stanislas, Montreal
Bring a problem. Leave with progress. The fastest way is LinkedIn — DMs open, response within a working day.