Case Studies

Selected projects from my work

A few examples of projects I've worked on. Some details are kept general to respect confidentiality.

Machine Learning E-commerce

Product Recommendation Engine

Challenge

A retail company needed to improve product discovery and increase sales. Their existing system relied on manual curation, which couldn't scale with a growing catalog of thousands of products.

Approach

Built a hybrid recommendation system combining collaborative filtering with content-based features. Deployed on AWS SageMaker for training, with production inference running on Kubernetes.

Stack

Python AWS SageMaker Pandas Kubernetes

Outcome

  • Increased conversions from recommendations
  • Higher engagement and click-through
  • Eliminated manual curation work
  • Auto-scales with traffic
Trading Systems Automation

Automated Stock Trading Platform

Challenge

Build a robust trading system for US equities that operates autonomously during market hours, executes trades based on predefined strategies, and handles real-time market data reliably.

Approach

Designed a complete automated platform using Alpaca for real-time market data and Interactive Brokers for execution. The system monitors positions and executes trades without manual intervention.

Stack

Python ib_insync Alpaca API asyncio

Outcome

  • Fully autonomous during market hours
  • Running live for over a year
  • Real-time market data processing
  • Reliable order execution
Mission Critical Events

High-Stakes Event Ticketing System

Challenge

A major event with 100,000+ tickets sold needed a custom seating system. Groups had to sit together, the system had to integrate with third-party ticketing platforms, and venue staff needed tools to handle issues on the spot. Zero margin for error.

Approach

Built a seat matching algorithm to keep groups together, integrated with external ticketing systems, and developed on-site tools for venue staff to resolve problems in real-time. Created an audit system to verify ticket consistency before the event.

Stack

Python API Integration Real-time Systems

Outcome

  • 100,000+ tickets processed
  • Groups seated together
  • Zero issues at the event
  • Full audit trail
Data Engineering E-commerce

Real-time Feature Store for Online Behaviour

Challenge

An e-commerce company needed to track user behaviour across their site in real-time to feed machine learning models. This required aggregating thousands of events from different sections of the site and making features available instantly for model training and inference.

Approach

Built a real-time feature storage system using Kafka to stream events from across the site. Implemented thousands of aggregations processed through Lambda functions, with Redis providing low-latency feature serving for ML models.

Stack

Python Kafka Redis AWS Lambda

Outcome

  • Real-time feature availability
  • Thousands of aggregations processed
  • Low-latency feature serving
  • ML-ready data pipeline
DevOps MLOps

ML Model Registry & Deployment Platform

Challenge

A client needed a way to manage dozens of ML models with different resource requirements. Deploying models was manual, scheduling training jobs was cumbersome, and there was no central place to track models or configure infrastructure.

Approach

Built a model registry platform where data scientists could register models and deploy them on demand. Each model could be scheduled independently with configurable instance resources. The platform handled data setup, training, and prediction with just a few clicks.

Stack

AWS Docker Kubernetes Prometheus Grafana

Outcome

  • Self-service model deployment
  • Flexible scheduling per model
  • Configurable resources per job
  • Data to prediction in clicks

Have a Similar Challenge?

Let's discuss how I can help with your project

Get in Touch