Backend and data engineer. I build services and pipelines that stay correct when data is late, duplicated, or malformed: Python and FastAPI on the serving side, PySpark and Delta Lake on the processing side, Kafka in between.
Now — Associate Software Engineer at Nagarro, Gurugram
Focus — Full stack development and data engineering
Open to — remote backend and data engineering roles
HVAC Telemetry Lakehouse
Aiven Kafka over TLS → Delta Lake medallion pipeline in PySpark → gold layer served by FastAPI on DuckDB, with Z-Ordering on the tables the API filters.
Code
Globe News Atlas
Ten public sources → bronze/silver/gold Parquet lake on Hugging Face → FastAPI → React/Three.js globe.
Live · Code
TaskFlow
Chat-native work OS spanning 62 Prisma models, with Socket.IO presence, threads, reactions, and one Docker service on Render.
Live · Code
apache/iceberg-python #4169 — merged. list-refs now emits one JSON object per ref, so CLI output parses cleanly downstream.
| # | Stage | Tooling |
|---|---|---|
| 01 | Sources | APIs · feeds · telemetry |
| 02 | Ingest | Python · Kafka · Azure Data Factory |
| 03 | Storage | Parquet · Delta Lake |
| 04 | Transform | PySpark · DuckDB · SQL |
| 05 | Quality | tests · quarantine · watermarks |
| 06 | Serve | FastAPI · Power BI |
Correctness before dashboards: I write the data-quality checks before the reporting layer.
- Sub-second OLAP without a warehouse: DuckDB + FastAPI
- Watermarks are a contract: validating 150+ PySpark streams
- How to integrate Clerk in a Next.js application
- dev.to/aayusbhat26
Python SQL PySpark Delta Lake Kafka DuckDB Azure Data Factory Databricks AWS FastAPI PostgreSQL Prisma TypeScript React Docker GitHub Actions
Portfolio · LinkedIn · LeetCode · work.aayush.bhat@gmail.com

