At EPAM Systems, I develop generative-AI and payments services for a US beauty retailer and a video streaming platform. I delivered short-term memory capabilities for an AI shopping assistant as part of a team initiative whose A/B test showed a statistically significant lift in conversion rate and revenue per visitor.
I built Databricks and PySpark pipelines that turn conversation and clickstream data into structured memories in PostgreSQL and pgvector, orchestrated with Azure Data Factory. I also integrated Azure OpenAI and LiteLLM and added structured logging and multi-region configuration.
For the streaming platform, I implemented Stripe SetupIntents, PaymentIntents and 3DS authentication with Zuora billing and Google Cloud Spanner migrations across Python, Java and PHP services. The work enabled subscription, invoice and VOD payments while deterring card-testing fraud through card-storage restrictions.
Earlier at EPAM Systems, I contributed GPU capabilities, PyTorch Docker images and a centralized Feature Store to a fashion retailer’s machine-learning platform on AWS and Kubernetes. At Rappi S.A.S., I worked on customer-support CRM services and BI pipelines; before that, I developed an airline CRM response model at BRM S.A.S. and an educational web application at Universe Horizons.
