Projects

Real systems I have built, for real businesses

Talk is easy. Here is what I have actually shipped.

OMERS · Financial Services

Reliable high-stakes pipeline architecture

The result

Risk & portfolio data

OMERS: Risk and portfolio data infrastructure

One of Canada's largest pension funds needed reliable financial data flowing from Bloomberg, MSCI, and SQL systems into a unified data warehouse that risk and portfolio teams could actually trust.

I built and maintained the ETL pipelines handling that data flow, created a Streamlit dashboard giving portfolio managers live visibility into risk data, and led a full migration of the orchestration layer from Apache Airflow to Prefect, all without disrupting daily operations.

The result: A pipeline architecture that has run reliably handling high-stakes financial data, with full visibility built in for the teams who depend on it.

  • Python
  • PostgreSQL
  • Bloomberg
  • Prefect
  • Streamlit
PSL Group · Healthcare

Full research day to a few minutes

The result

AI research assistant

PSL Group: AI research assistant for medical teams

A healthcare AI startup had a research team spending full days manually reading through clinical literature to find relevant information.

I built a serverless LLM application on AWS that ingests documents, extracts structured insights using GCP Vertex AI and OpenAI, and surfaces findings through a clean interface in seconds instead of hours.

The result: What used to take a full research day now takes a few minutes.

  • AWS
  • Vertex AI
  • OpenAI
  • Python
Marketing by Professionals · Agency

Team focused on client relationships

The result

Full marketing automation

Marketing by Professionals: Full AI marketing automation engine

A marketing agency was running their entire client acquisition pipeline manually: outreach, follow-ups, proposals, and lead questions, all by hand.

I rebuilt the system from the ground up: n8n-powered outreach automation, an AI-driven sales funnel, a proposal generation system that produces a customized proposal in under two minutes, and AI chatbots that qualify leads before handing off to the team.

The result: The team stopped spending their time on tasks a system could do better, and started focusing on actual client relationships.

  • n8n
  • OpenAI
  • Python
  • React
Scoodia · SaaS

Hours to under five minutes

The result

Intelligent request routing

Scoodia: Intelligent request routing system

A growing SaaS company had an operations team drowning in customer requests coming through email, forms, and Slack, with no consistent way to prioritize or route them.

I built an AI routing system using Claude and LangChain that reads every incoming request, understands intent and urgency using natural language, pulls relevant account context automatically, and either resolves it on the spot or hands it off with a full briefing already prepared.

The result: Average response time dropped from several hours to under five minutes.

  • Claude
  • LangChain
  • Python
TMRC · Enterprise

Two live production platforms

The result

HR portal & e-commerce

TMRC: HR portal and e-commerce platform

An enterprise client needed two separate production systems: a Django and React HR and payroll portal with live integrations to Microsoft Business Central, and an international e-commerce platform with the same backend integration.

I built both, handling background task processing with Celery and Redis to keep the systems responsive under real-world load.

The result: Two live production platforms still supporting daily business operations.

  • Django
  • React
  • Celery
  • Redis

See something here that looks like your problem? Get in touch and let's figure out what the version of this looks like for your business.