Hit me up for enquiries —WhatsAppEmailCallback
Backend & data infrastructure engineer at Happy-Place (ISPG). I design streaming pipelines, lakehouses and the Python services around them — and run them in production on Kubernetes.
CDC → stream → medallion lakehouse → query
Debezium captures changes from any source database — MongoDB in production. Delta Lake on MinIO, orchestrated by Airflow, with each tenant isolated on Pulsar by its own credentials and policies.
2021 — now
Built from concept to production. A code-driven control plane configures the whole Kubernetes cluster with four levels of data separation — conglomerate, organization, project, environment. A no-code builder generates PySpark jobs and Airflow DAGs, feeding a CDC-driven Delta Lake medallion lakehouse.
Python · FastAPI · PySpark · Airflow · Debezium · Pulsar · Delta Lake · Trino · Kubernetes · Argo CD
Read the case study →The engine behind how people find things on a marketplace built for commerce across many countries. Heavily nested product documents, hybrid keyword and vector retrieval, and the MLOps around the models that rank them.
Go · MongoDB · OpenSearch · vector embeddings · Qwen · MLOps
Standalone B2C / B2B / enterprise store for a PC and server retailer, with a WhatsApp Business tool and helpdesk.
Golf booking platform designed for many countries, many timezones and thousands of concurrent users.
Food delivery backend scaled to three countries and 5,000+ active users.
Self-initiated penetration test of internal deployments; found critical CVEs including RCE in Kafka UI.
How we stream MongoDB changes through Debezium Server and Apache Pulsar into a Delta Lake lakehouse, and the silent failures we had to turn into loud ones.
Five years across product companies and agencies in India and the UAE — moving steadily from APIs toward the data underneath them.
Enquiries
Hit me up. Message me on WhatsApp, send an email, or leave your details and I'll call you back.
© 2026 George Thomas · georgev.dev