LLM Orchestration
Designing multi-agent pipelines with reasoning loops, tool use, and memory — built on top of OpenAI, Anthropic, and Ollama models.
I blend applied machine learning, AI agents, and full-stack engineering to turn complex ideas into production systems — from enterprise cloud platforms to self-hosted infrastructure.

I'm a Python Developer & AI Engineer specialising in designing and deploying intelligent agent systems, generative AI solutions, and end-to-end data engineering pipelines.
At Carnival Cruise Line and previously at Qualex Consulting Services, I've integrated AWS, Databricks, and cutting-edge LLMs to deliver enterprise-grade AI products that operate at scale.
Beyond enterprise work, I run a self-hosted home server — deploying AI models, automation workflows, and multiple web apps entirely from my own hardware.
Designing multi-agent pipelines with reasoning loops, tool use, and memory — built on top of OpenAI, Anthropic, and Ollama models.
Building no-code/low-code AI automation pipelines with n8n — connecting APIs, AI models, and data sources into production workflows.
Provisioning scalable ML infrastructure on AWS — SageMaker, Lambda, EC2 — alongside Databricks for large-scale data processing.
Training, fine-tuning, and deploying deep learning models with TensorFlow and PyTorch for classification, prediction, and NLP tasks.
Shipping production-ready REST and async APIs in Flask & Django that expose AI capabilities — with clean, scalable backend architecture.
Running local AI models via Ollama, reverse-proxied through nginx, automated via n8n — full AI stack on personal hardware.
AI-powered wardrobe cataloger and personal stylist — cross-platform React Native + Expo app with a FastAPI backend. Claude AI vision-tags clothing from photos, generates outfit combinations, and supports Stripe subscriptions. Full production QA audit and 33-issue bug-fix sprint delivered.
Self-hosted homelab AI platform running 24/7 — nginx reverse proxy, n8n automation, local Ollama LLMs, and automated backups. DURYO Agent continuously monitors services, summarises incidents with AI, and fires alerts to keep the stack healthy.
End-to-end ML study of Miami's housing market — curated datasets, feature engineering, regression and ensemble models, plus Jupyter notebooks and visualisations that surface market signals for non-technical stakeholders.
Retrieval-Augmented Generation chatbot that ingests PDFs and answers questions using vector search and LLM reasoning — supports multi-document context, semantic chunking, and streaming responses.
Full MLOps pipeline for fine-tuning and deploying custom LLMs on AWS SageMaker — includes data preparation, training job automation, inference endpoint deployment, and shell scripting for end-to-end reproducibility.
AI SaaS that auto-generates and schedules LinkedIn posts — LLM-powered content engine, Docker-containerised backend, web dashboard for post management, and n8n workflow orchestrator for publish scheduling.
Conversational AI for hiring teams — screens candidate FAQs, captures structured applicant data, routes qualified leads to recruiters, and reduces manual back-and-forth across the full recruitment funnel.
Premium e-commerce site for the Tesla EXPLR9 rugged phone — Three.js scroll-driven 3D phone animation, orbital ring effects, product gallery with real images, YouTube integration, and full Next.js commerce architecture.
Full Next.js clone of luxy-events.com — production-grade event company website featuring Diamond Lines partnership content, partner badges, meal-prep section, and polished multi-page architecture.
Futuristic tech product landing page built with Orbitron + Inter typography, animated sections, and bundled n8n remote-access scripts for AI automation setup — showcasing speculative tech branding.
Full-service smoke shop website for Weedzies Miami in Little Haiti — branded with Monoton + Bungee typography, neon-pink theme, age-gate landing, product category pages, Google Maps embed, and live SEO-optimised meta for local search.

Amazon Web Services · Associate · Active

Google Cloud · Professional · Active

Google Cloud · Associate · Active

Microsoft Certified · Associate · DP-100

Oracle Certified · Associate · 1Z0-1085-25

Databricks Academy · Accreditation · Active

Microsoft Certified · DP-420
DeepLearning.AI · Short course
Custom LLM agents, RAG pipelines, tool-use orchestration — built to run in production.
Auto-generate social posts, ad copy, and campaigns with LLM content engines + n8n scheduling.
Plug ChatGPT, Claude, or Ollama into your existing business tools, CRMs, and workflows.
n8n, webhooks, and AI triggers that eliminate repetitive tasks across your entire operation.
ChatGPT-driven keyword research, auto-generated meta tags, SEO content writing, and competitor gap analysis at scale.
Fast, responsive, SEO-optimised sites — from landing pages to full multi-page business sites.
Full-stack web apps with auth, billing, dashboards, and API backends — shipped to production.
Online stores with product pages, cart, Stripe checkout, and inventory — ready to sell.
REST APIs in Flask, Django, or Next.js — clean architecture, documented, and scalable.
AWS, GCP, and Azure architecture, provisioning, cost optimisation, and ongoing ops.
Homelab or on-prem server setup — nginx, Docker, VPN, reverse proxies, and automation.
SQL schema design, query tuning, data pipelines, and ETL from raw sources to insight.
Secure network design, VLANs, firewall config, Cloudflare tunnels, and remote access.
Executive dashboards, KPI tracking, and automated reports that surface what actually matters.
Architecture reviews, tech stack decisions, and strategic advisory for startups and SMBs.
On-page SEO, Google Business, structured data, page speed, backlinks, and local search — full visibility strategy.
Bug triage, documentation, team onboarding, and clean handoff for any tech system.
Available for projects & roles
Need an AI engineer, data platform, or custom agent system? I offer end-to-end services from architecture to deployment — remote, hybrid, or Miami on-site.
Client-side simulation
Mimics retrieve → cite → answer. Not connected to an LLM or database.